From b47edca6bed5814e93a0422fc29279587de0d553 Mon Sep 17 00:00:00 2001 From: Dalton Hubble Date: Sat, 19 Mar 2022 17:04:14 -0700 Subject: [PATCH] Refresh Prometheus rules and Grafana dashboards * Update Prometheus rules and Grafana dashboards * Add new networking dashboards --- CHANGES.md | 1 + addons/grafana/dashboards-etcd.yaml | 153 +- addons/grafana/dashboards-k8s-network.yaml | 7644 +++++++++++++++++ addons/grafana/dashboards-k8s-nodes.yaml | 4581 +++++----- .../grafana/dashboards-k8s-resources-1.yaml | 1589 +++- .../grafana/dashboards-k8s-resources-2.yaml | 1467 +++- addons/grafana/dashboards-k8s.yaml | 1246 +-- addons/grafana/dashboards-node-exporter.yaml | 32 +- addons/grafana/dashboards-prom.yaml | 46 +- addons/grafana/deployment.yaml | 5 + addons/prometheus/rules.yaml | 537 +- 11 files changed, 12898 insertions(+), 4403 deletions(-) create mode 100644 addons/grafana/dashboards-k8s-network.yaml diff --git a/CHANGES.md b/CHANGES.md index 1e0a4ff2..effe1728 100644 --- a/CHANGES.md +++ b/CHANGES.md @@ -13,6 +13,7 @@ Notable changes between versions. ### Addons +* Refresh Prometheus rules and Grafana dashboards ([#1136](https://github.com/poseidon/typhoon/pull/1136)) * Update nginx-ingress from v1.1.1 to [v1.1.2](https://github.com/kubernetes/ingress-nginx/releases/tag/controller-v1.1.2) * Update Prometheus from v2.33.3 to [v2.33.5](https://github.com/prometheus/prometheus/releases/tag/v2.33.5) * Update Grafana from v8.4.1 to [v8.4.3](https://github.com/grafana/grafana/releases/tag/v8.4.3) diff --git a/addons/grafana/dashboards-etcd.yaml b/addons/grafana/dashboards-etcd.yaml index 6fa287eb..b5b97c81 100644 --- a/addons/grafana/dashboards-etcd.yaml +++ b/addons/grafana/dashboards-etcd.yaml @@ -140,7 +140,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(rate(grpc_server_started_total{job=\"$cluster\",grpc_type=\"unary\"}[5m]))", + "expr": "sum(rate(grpc_server_started_total{job=\"$cluster\",grpc_type=\"unary\"}[$__rate_interval]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "RPC Rate", @@ -149,7 +149,7 @@ data: "step": 2 }, { - "expr": "sum(rate(grpc_server_handled_total{job=\"$cluster\",grpc_type=\"unary\",grpc_code!=\"OK\"}[5m]))", + "expr": "sum(rate(grpc_server_handled_total{job=\"$cluster\",grpc_type=\"unary\",grpc_code=~\"Unknown|FailedPrecondition|ResourceExhausted|Internal|Unavailable|DataLoss|DeadlineExceeded\"}[$__rate_interval]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "RPC Failed Rate", @@ -430,7 +430,7 @@ data: "steppedLine": true, "targets": [ { - "expr": "histogram_quantile(0.99, sum(rate(etcd_disk_wal_fsync_duration_seconds_bucket{job=\"$cluster\"}[5m])) by (instance, le))", + "expr": "histogram_quantile(0.99, sum(rate(etcd_disk_wal_fsync_duration_seconds_bucket{job=\"$cluster\"}[$__rate_interval])) by (instance, le))", "hide": false, "intervalFactor": 2, "legendFormat": "{{instance}} WAL fsync", @@ -439,7 +439,7 @@ data: "step": 4 }, { - "expr": "histogram_quantile(0.99, sum(rate(etcd_disk_backend_commit_duration_seconds_bucket{job=\"$cluster\"}[5m])) by (instance, le))", + "expr": "histogram_quantile(0.99, sum(rate(etcd_disk_backend_commit_duration_seconds_bucket{job=\"$cluster\"}[$__rate_interval])) by (instance, le))", "intervalFactor": 2, "legendFormat": "{{instance}} DB fsync", "metric": "etcd_disk_backend_commit_duration_seconds_bucket", @@ -617,7 +617,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "rate(etcd_network_client_grpc_received_bytes_total{job=\"$cluster\"}[5m])", + "expr": "rate(etcd_network_client_grpc_received_bytes_total{job=\"$cluster\"}[$__rate_interval])", "intervalFactor": 2, "legendFormat": "{{instance}} Client Traffic In", "metric": "etcd_network_client_grpc_received_bytes_total", @@ -703,7 +703,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "rate(etcd_network_client_grpc_sent_bytes_total{job=\"$cluster\"}[5m])", + "expr": "rate(etcd_network_client_grpc_sent_bytes_total{job=\"$cluster\"}[$__rate_interval])", "intervalFactor": 2, "legendFormat": "{{instance}} Client Traffic Out", "metric": "etcd_network_client_grpc_sent_bytes_total", @@ -789,7 +789,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(rate(etcd_network_peer_received_bytes_total{job=\"$cluster\"}[5m])) by (instance)", + "expr": "sum(rate(etcd_network_peer_received_bytes_total{job=\"$cluster\"}[$__rate_interval])) by (instance)", "intervalFactor": 2, "legendFormat": "{{instance}} Peer Traffic In", "metric": "etcd_network_peer_received_bytes_total", @@ -878,7 +878,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(rate(etcd_network_peer_sent_bytes_total{job=\"$cluster\"}[5m])) by (instance)", + "expr": "sum(rate(etcd_network_peer_sent_bytes_total{job=\"$cluster\"}[$__rate_interval])) by (instance)", "hide": false, "interval": "", "intervalFactor": 2, @@ -972,7 +972,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(rate(etcd_server_proposals_failed_total{job=\"$cluster\"}[5m]))", + "expr": "sum(rate(etcd_server_proposals_failed_total{job=\"$cluster\"}[$__rate_interval]))", "intervalFactor": 2, "legendFormat": "Proposal Failure Rate", "metric": "etcd_server_proposals_failed_total", @@ -988,7 +988,7 @@ data: "step": 2 }, { - "expr": "sum(rate(etcd_server_proposals_committed_total{job=\"$cluster\"}[5m]))", + "expr": "sum(rate(etcd_server_proposals_committed_total{job=\"$cluster\"}[$__rate_interval]))", "intervalFactor": 2, "legendFormat": "Proposal Commit Rate", "metric": "etcd_server_proposals_committed_total", @@ -996,7 +996,7 @@ data: "step": 2 }, { - "expr": "sum(rate(etcd_server_proposals_applied_total{job=\"$cluster\"}[5m]))", + "expr": "sum(rate(etcd_server_proposals_applied_total{job=\"$cluster\"}[$__rate_interval]))", "intervalFactor": 2, "legendFormat": "Proposal Apply Rate", "refId": "D", @@ -1131,6 +1131,131 @@ data: "show": true } ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "decimals": 0, + "editable": true, + "error": false, + "fieldConfig": { + "defaults": { + "custom": { + + } + }, + "overrides": [ + + ] + }, + "fill": 0, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 28 + }, + "hiddenSeries": false, + "id": 42, + "isNew": true, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": false, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "nullPointMode": "connected", + "options": { + "alertThreshold": true + }, + "percentage": false, + "pluginVersion": "7.4.3", + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.99, sum by (instance, le) (rate(etcd_network_peer_round_trip_time_seconds_bucket{job=\"$cluster\"}[$__rate_interval])))", + "interval": "", + "intervalFactor": 2, + "legendFormat": "{{instance}} Peer round trip time", + "metric": "etcd_network_peer_round_trip_time_seconds_bucket", + "refId": "A", + "step": 2 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeRegions": [ + + ], + "timeShift": null, + "title": "Peer round trip time", + "tooltip": { + "msResolution": false, + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "$$hashKey": "object:925", + "decimals": null, + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "$$hashKey": "object:926", + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ], + "yaxis": { + "align": false, + "alignLevel": null + } } ], "title": "New row" @@ -1140,7 +1265,7 @@ data: "sharedCrosshair": false, "style": "dark", "tags": [ - + "etcd-mixin" ], "templating": { "list": [ @@ -1150,7 +1275,7 @@ data: "value": "Prometheus" }, "hide": 0, - "label": null, + "label": "Data Source", "name": "datasource", "options": [ @@ -1176,7 +1301,7 @@ data: ], "query": "label_values(etcd_server_has_leader, job)", - "refresh": 1, + "refresh": 2, "regex": "", "sort": 2, "tagValuesQuery": "", diff --git a/addons/grafana/dashboards-k8s-network.yaml b/addons/grafana/dashboards-k8s-network.yaml new file mode 100644 index 00000000..13733a0a --- /dev/null +++ b/addons/grafana/dashboards-k8s-network.yaml @@ -0,0 +1,7644 @@ +apiVersion: v1 +data: + cluster-total.json: |- + { + "__inputs": [ + + ], + "__requires": [ + + ], + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": "-- Grafana --", + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "type": "dashboard" + } + ] + }, + "editable": true, + "gnetId": null, + "graphTooltip": 0, + "hideControls": false, + "id": null, + "links": [ + + ], + "panels": [ + { + "collapse": false, + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 0 + }, + "id": 2, + "panels": [ + + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Current Bandwidth", + "titleSize": "h6", + "type": "row" + }, + { + "aliasColors": { + + }, + "bars": true, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 1 + }, + "id": 3, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": false, + "linewidth": 1, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "null", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 24, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\",namespace=~\".+\"}[$interval:$resolution])) by (namespace))", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{namespace}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Current Rate of Bytes Received", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "series", + "name": null, + "show": false, + "values": [ + "current" + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": true, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 1 + }, + "id": 4, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": false, + "linewidth": 1, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "null", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 24, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\",namespace=~\".+\"}[$interval:$resolution])) by (namespace))", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{namespace}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Current Rate of Bytes Transmitted", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "series", + "name": null, + "show": false, + "values": [ + "current" + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "columns": [ + { + "text": "Time", + "value": "Time" + }, + { + "text": "Value #A", + "value": "Value #A" + }, + { + "text": "Value #B", + "value": "Value #B" + }, + { + "text": "Value #C", + "value": "Value #C" + }, + { + "text": "Value #D", + "value": "Value #D" + }, + { + "text": "Value #E", + "value": "Value #E" + }, + { + "text": "Value #F", + "value": "Value #F" + }, + { + "text": "Value #G", + "value": "Value #G" + }, + { + "text": "Value #H", + "value": "Value #H" + }, + { + "text": "namespace", + "value": "namespace" + } + ], + "datasource": "$datasource", + "fill": 1, + "fontSize": "90%", + "gridPos": { + "h": 9, + "w": 24, + "x": 0, + "y": 10 + }, + "id": 5, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "null as zero", + "renderer": "flot", + "scroll": true, + "showHeader": true, + "sort": { + "col": 0, + "desc": false + }, + "spaceLength": 10, + "span": 24, + "styles": [ + { + "alias": "Time", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Time", + "thresholds": [ + + ], + "type": "hidden", + "unit": "short" + }, + { + "alias": "Current Bandwidth Received", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #A", + "thresholds": [ + + ], + "type": "number", + "unit": "Bps" + }, + { + "alias": "Current Bandwidth Transmitted", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #B", + "thresholds": [ + + ], + "type": "number", + "unit": "Bps" + }, + { + "alias": "Average Bandwidth Received", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #C", + "thresholds": [ + + ], + "type": "number", + "unit": "Bps" + }, + { + "alias": "Average Bandwidth Transmitted", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #D", + "thresholds": [ + + ], + "type": "number", + "unit": "Bps" + }, + { + "alias": "Rate of Received Packets", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #E", + "thresholds": [ + + ], + "type": "number", + "unit": "pps" + }, + { + "alias": "Rate of Transmitted Packets", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #F", + "thresholds": [ + + ], + "type": "number", + "unit": "pps" + }, + { + "alias": "Rate of Received Packets Dropped", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #G", + "thresholds": [ + + ], + "type": "number", + "unit": "pps" + }, + { + "alias": "Rate of Transmitted Packets Dropped", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #H", + "thresholds": [ + + ], + "type": "number", + "unit": "pps" + }, + { + "alias": "Namespace", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": true, + "linkTooltip": "Drill down", + "linkUrl": "d/8b7a8b326d7a6f1f04244066368c67af/kubernetes-networking-namespace-pods?orgId=1&refresh=30s&var-namespace=$__cell", + "pattern": "namespace", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + } + ], + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\",namespace=~\".+\"}[$interval:$resolution])) by (namespace))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "A", + "step": 10 + }, + { + "expr": "sort_desc(sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\",namespace=~\".+\"}[$interval:$resolution])) by (namespace))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "B", + "step": 10 + }, + { + "expr": "sort_desc(avg(irate(container_network_receive_bytes_total{cluster=\"$cluster\",namespace=~\".+\"}[$interval:$resolution])) by (namespace))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "C", + "step": 10 + }, + { + "expr": "sort_desc(avg(irate(container_network_transmit_bytes_total{cluster=\"$cluster\",namespace=~\".+\"}[$interval:$resolution])) by (namespace))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "D", + "step": 10 + }, + { + "expr": "sort_desc(sum(irate(container_network_receive_packets_total{cluster=\"$cluster\",namespace=~\".+\"}[$interval:$resolution])) by (namespace))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "E", + "step": 10 + }, + { + "expr": "sort_desc(sum(irate(container_network_transmit_packets_total{cluster=\"$cluster\",namespace=~\".+\"}[$interval:$resolution])) by (namespace))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "F", + "step": 10 + }, + { + "expr": "sort_desc(sum(irate(container_network_receive_packets_dropped_total{cluster=\"$cluster\",namespace=~\".+\"}[$interval:$resolution])) by (namespace))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "G", + "step": 10 + }, + { + "expr": "sort_desc(sum(irate(container_network_transmit_packets_dropped_total{cluster=\"$cluster\",namespace=~\".+\"}[$interval:$resolution])) by (namespace))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "H", + "step": 10 + } + ], + "timeFrom": null, + "timeShift": null, + "title": "Current Status", + "type": "table" + }, + { + "collapse": true, + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 10 + }, + "id": 6, + "panels": [ + { + "aliasColors": { + + }, + "bars": true, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 11 + }, + "id": 7, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": false, + "linewidth": 1, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "null", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 24, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(avg(irate(container_network_receive_bytes_total{cluster=\"$cluster\",namespace=~\".+\"}[$interval:$resolution])) by (namespace))", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{namespace}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Average Rate of Bytes Received", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "series", + "name": null, + "show": false, + "values": [ + "current" + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": true, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 11 + }, + "id": 8, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": false, + "linewidth": 1, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "null", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 24, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(avg(irate(container_network_transmit_bytes_total{cluster=\"$cluster\",namespace=~\".+\"}[$interval:$resolution])) by (namespace))", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{namespace}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Average Rate of Bytes Transmitted", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "series", + "name": null, + "show": false, + "values": [ + "current" + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Average Bandwidth", + "titleSize": "h6", + "type": "row" + }, + { + "collapse": false, + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 11 + }, + "id": 9, + "panels": [ + + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Bandwidth History", + "titleSize": "h6", + "type": "row" + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 24, + "x": 0, + "y": 12 + }, + "id": 10, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": true, + "min": true, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 24, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\",namespace=~\".+\"}[$interval:$resolution])) by (namespace))", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{namespace}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Receive Bandwidth", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 24, + "x": 0, + "y": 21 + }, + "id": 11, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": true, + "min": true, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 24, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\",namespace=~\".+\"}[$interval:$resolution])) by (namespace))", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{namespace}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Transmit Bandwidth", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "collapse": true, + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 30 + }, + "id": 12, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 24, + "x": 0, + "y": 31 + }, + "id": 13, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": true, + "min": true, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 24, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_receive_packets_total{cluster=\"$cluster\",namespace=~\".+\"}[$interval:$resolution])) by (namespace))", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{namespace}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of Received Packets", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 24, + "x": 0, + "y": 40 + }, + "id": 14, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": true, + "min": true, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 24, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_transmit_packets_total{cluster=\"$cluster\",namespace=~\".+\"}[$interval:$resolution])) by (namespace))", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{namespace}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of Transmitted Packets", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Packets", + "titleSize": "h6", + "type": "row" + }, + { + "collapse": true, + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 31 + }, + "id": 15, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 24, + "x": 0, + "y": 50 + }, + "id": 16, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": true, + "min": true, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 24, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_receive_packets_dropped_total{cluster=\"$cluster\",namespace=~\".+\"}[$interval:$resolution])) by (namespace))", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{namespace}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of Received Packets Dropped", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 24, + "x": 0, + "y": 59 + }, + "id": 17, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": true, + "min": true, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 24, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_transmit_packets_dropped_total{cluster=\"$cluster\",namespace=~\".+\"}[$interval:$resolution])) by (namespace))", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{namespace}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of Transmitted Packets Dropped", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 24, + "x": 0, + "y": 59 + }, + "id": 18, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": true, + "min": true, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [ + { + "targetBlank": true, + "title": "What is TCP Retransmit?", + "url": "https://accedian.com/enterprises/blog/network-packet-loss-retransmissions-and-duplicate-acknowledgements/" + } + ], + "minSpan": 24, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 24, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(rate(node_netstat_Tcp_RetransSegs{cluster=\"$cluster\"}[$interval:$resolution]) / rate(node_netstat_Tcp_OutSegs{cluster=\"$cluster\"}[$interval:$resolution])) by (instance))", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{instance}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of TCP Retransmits out of all sent segments", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "percentunit", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "percentunit", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 24, + "x": 0, + "y": 59 + }, + "id": 19, + "legend": { + "alignAsTable": true, + "avg": true, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": true, + "min": true, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 2, + "links": [ + { + "targetBlank": true, + "title": "Why monitor SYN retransmits?", + "url": "https://github.com/prometheus/node_exporter/issues/1023#issuecomment-408128365" + } + ], + "minSpan": 24, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 24, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(rate(node_netstat_TcpExt_TCPSynRetrans{cluster=\"$cluster\"}[$interval:$resolution]) / rate(node_netstat_Tcp_RetransSegs{cluster=\"$cluster\"}[$interval:$resolution])) by (instance))", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{instance}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of TCP SYN Retransmits out of all retransmits", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "percentunit", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "percentunit", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Errors", + "titleSize": "h6", + "type": "row" + } + ], + "refresh": "10s", + "rows": [ + + ], + "schemaVersion": 18, + "style": "dark", + "tags": [ + "kubernetes-mixin" + ], + "templating": { + "list": [ + { + "allValue": null, + "auto": false, + "auto_count": 30, + "auto_min": "10s", + "current": { + "text": "5m", + "value": "5m" + }, + "datasource": "$datasource", + "hide": 0, + "includeAll": false, + "label": null, + "multi": false, + "name": "resolution", + "options": [ + { + "selected": false, + "text": "30s", + "value": "30s" + }, + { + "selected": true, + "text": "5m", + "value": "5m" + }, + { + "selected": false, + "text": "1h", + "value": "1h" + } + ], + "query": "30s,5m,1h", + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "interval", + "useTags": false + }, + { + "allValue": null, + "auto": false, + "auto_count": 30, + "auto_min": "10s", + "current": { + "text": "5m", + "value": "5m" + }, + "datasource": "$datasource", + "hide": 2, + "includeAll": false, + "label": null, + "multi": false, + "name": "interval", + "options": [ + { + "selected": true, + "text": "4h", + "value": "4h" + } + ], + "query": "4h", + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "interval", + "useTags": false + }, + { + "current": { + "text": "default", + "value": "default" + }, + "hide": 0, + "label": "Data Source", + "name": "datasource", + "options": [ + + ], + "query": "prometheus", + "refresh": 1, + "regex": "", + "type": "datasource" + }, + { + "allValue": null, + "current": { + + }, + "datasource": "$datasource", + "hide": 2, + "includeAll": false, + "label": null, + "multi": false, + "name": "cluster", + "options": [ + + ], + "query": "label_values(up{job=\"kubernetes-cadvisor\"}, cluster)", + "refresh": 2, + "regex": "", + "sort": 0, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + } + ] + }, + "time": { + "from": "now-1h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "5s", + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ], + "time_options": [ + "5m", + "15m", + "1h", + "6h", + "12h", + "24h", + "2d", + "7d", + "30d" + ] + }, + "timezone": "UTC", + "title": "Kubernetes / Networking / Cluster", + "uid": "ff635a025bcfea7bc3dd4f508990a3e9", + "version": 0 + } + namespace-by-pod.json: |- + { + "__inputs": [ + + ], + "__requires": [ + + ], + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": "-- Grafana --", + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "type": "dashboard" + } + ] + }, + "editable": true, + "gnetId": null, + "graphTooltip": 0, + "hideControls": false, + "id": null, + "links": [ + + ], + "panels": [ + { + "collapse": false, + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 0 + }, + "id": 2, + "panels": [ + + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Current Bandwidth", + "titleSize": "h6", + "type": "row" + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": [ + "#299c46", + "rgba(237, 129, 40, 0.89)", + "#d44a3a" + ], + "datasource": "$datasource", + "decimals": 0, + "format": "time_series", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 1 + }, + "height": 9, + "id": 3, + "interval": null, + "links": [ + + ], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "minSpan": 12, + "nullPointMode": "connected", + "nullText": null, + "options": { + "fieldOptions": { + "calcs": [ + "last" + ], + "defaults": { + "max": 10000000000, + "min": 0, + "title": "$namespace", + "unit": "Bps" + }, + "mappings": [ + + ], + "override": { + + }, + "thresholds": [ + { + "color": "dark-green", + "index": 0, + "value": null + }, + { + "color": "dark-yellow", + "index": 1, + "value": 5000000000 + }, + { + "color": "dark-red", + "index": 2, + "value": 7000000000 + } + ], + "values": false + } + }, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 12, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution]))", + "format": "time_series", + "instant": null, + "intervalFactor": 1, + "legendFormat": "", + "refId": "A" + } + ], + "thresholds": "", + "timeFrom": null, + "timeShift": null, + "title": "Current Rate of Bytes Received", + "type": "gauge", + "valueFontSize": "80%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": [ + "#299c46", + "rgba(237, 129, 40, 0.89)", + "#d44a3a" + ], + "datasource": "$datasource", + "decimals": 0, + "format": "time_series", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 1 + }, + "height": 9, + "id": 4, + "interval": null, + "links": [ + + ], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "minSpan": 12, + "nullPointMode": "connected", + "nullText": null, + "options": { + "fieldOptions": { + "calcs": [ + "last" + ], + "defaults": { + "max": 10000000000, + "min": 0, + "title": "$namespace", + "unit": "Bps" + }, + "mappings": [ + + ], + "override": { + + }, + "thresholds": [ + { + "color": "dark-green", + "index": 0, + "value": null + }, + { + "color": "dark-yellow", + "index": 1, + "value": 5000000000 + }, + { + "color": "dark-red", + "index": 2, + "value": 7000000000 + } + ], + "values": false + } + }, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 12, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution]))", + "format": "time_series", + "instant": null, + "intervalFactor": 1, + "legendFormat": "", + "refId": "A" + } + ], + "thresholds": "", + "timeFrom": null, + "timeShift": null, + "title": "Current Rate of Bytes Transmitted", + "type": "gauge", + "valueFontSize": "80%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + }, + { + "columns": [ + { + "text": "Time", + "value": "Time" + }, + { + "text": "Value #A", + "value": "Value #A" + }, + { + "text": "Value #B", + "value": "Value #B" + }, + { + "text": "Value #C", + "value": "Value #C" + }, + { + "text": "Value #D", + "value": "Value #D" + }, + { + "text": "Value #E", + "value": "Value #E" + }, + { + "text": "Value #F", + "value": "Value #F" + }, + { + "text": "pod", + "value": "pod" + } + ], + "datasource": "$datasource", + "fill": 1, + "fontSize": "100%", + "gridPos": { + "h": 9, + "w": 24, + "x": 0, + "y": 10 + }, + "id": 5, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "null as zero", + "renderer": "flot", + "scroll": true, + "showHeader": true, + "sort": { + "col": 0, + "desc": false + }, + "spaceLength": 10, + "span": 24, + "styles": [ + { + "alias": "Time", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Time", + "thresholds": [ + + ], + "type": "hidden", + "unit": "short" + }, + { + "alias": "Bandwidth Received", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #A", + "thresholds": [ + + ], + "type": "number", + "unit": "Bps" + }, + { + "alias": "Bandwidth Transmitted", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #B", + "thresholds": [ + + ], + "type": "number", + "unit": "Bps" + }, + { + "alias": "Rate of Received Packets", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #C", + "thresholds": [ + + ], + "type": "number", + "unit": "pps" + }, + { + "alias": "Rate of Transmitted Packets", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #D", + "thresholds": [ + + ], + "type": "number", + "unit": "pps" + }, + { + "alias": "Rate of Received Packets Dropped", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #E", + "thresholds": [ + + ], + "type": "number", + "unit": "pps" + }, + { + "alias": "Rate of Transmitted Packets Dropped", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #F", + "thresholds": [ + + ], + "type": "number", + "unit": "pps" + }, + { + "alias": "Pod", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": true, + "linkTooltip": "Drill down", + "linkUrl": "d/7a18067ce943a40ae25454675c19ff5c/kubernetes-networking-pod?orgId=1&refresh=30s&var-namespace=$namespace&var-pod=$__cell", + "pattern": "pod", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + } + ], + "targets": [ + { + "expr": "sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])) by (pod)", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "A", + "step": 10 + }, + { + "expr": "sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])) by (pod)", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "B", + "step": 10 + }, + { + "expr": "sum(irate(container_network_receive_packets_total{cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])) by (pod)", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "C", + "step": 10 + }, + { + "expr": "sum(irate(container_network_transmit_packets_total{cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])) by (pod)", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "D", + "step": 10 + }, + { + "expr": "sum(irate(container_network_receive_packets_dropped_total{cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])) by (pod)", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "E", + "step": 10 + }, + { + "expr": "sum(irate(container_network_transmit_packets_dropped_total{cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])) by (pod)", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "F", + "step": 10 + } + ], + "timeFrom": null, + "timeShift": null, + "title": "Current Status", + "type": "table" + }, + { + "collapse": false, + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 19 + }, + "id": 6, + "panels": [ + + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Bandwidth", + "titleSize": "h6", + "type": "row" + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 20 + }, + "id": 7, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])) by (pod)", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{pod}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Receive Bandwidth", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 20 + }, + "id": 8, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])) by (pod)", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{pod}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Transmit Bandwidth", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "collapse": true, + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 29 + }, + "id": 9, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 10, + "w": 12, + "x": 0, + "y": 30 + }, + "id": 10, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum(irate(container_network_receive_packets_total{cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])) by (pod)", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{pod}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of Received Packets", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 10, + "w": 12, + "x": 12, + "y": 30 + }, + "id": 11, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum(irate(container_network_transmit_packets_total{cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])) by (pod)", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{pod}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of Transmitted Packets", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Packets", + "titleSize": "h6", + "type": "row" + }, + { + "collapse": true, + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 30 + }, + "id": 12, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 10, + "w": 12, + "x": 0, + "y": 40 + }, + "id": 13, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum(irate(container_network_receive_packets_dropped_total{cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])) by (pod)", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{pod}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of Received Packets Dropped", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 10, + "w": 12, + "x": 12, + "y": 40 + }, + "id": 14, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum(irate(container_network_transmit_packets_dropped_total{cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])) by (pod)", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{pod}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of Transmitted Packets Dropped", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Errors", + "titleSize": "h6", + "type": "row" + } + ], + "refresh": "10s", + "rows": [ + + ], + "schemaVersion": 18, + "style": "dark", + "tags": [ + "kubernetes-mixin" + ], + "templating": { + "list": [ + { + "current": { + "text": "default", + "value": "default" + }, + "hide": 0, + "label": "Data Source", + "name": "datasource", + "options": [ + + ], + "query": "prometheus", + "refresh": 1, + "regex": "", + "type": "datasource" + }, + { + "allValue": null, + "current": { + + }, + "datasource": "$datasource", + "hide": 2, + "includeAll": false, + "label": null, + "multi": false, + "name": "cluster", + "options": [ + + ], + "query": "label_values(up{job=\"kubernetes-cadvisor\"}, cluster)", + "refresh": 2, + "regex": "", + "sort": 0, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": ".+", + "auto": false, + "auto_count": 30, + "auto_min": "10s", + "current": { + "text": "kube-system", + "value": "kube-system" + }, + "datasource": "$datasource", + "definition": "label_values(container_network_receive_packets_total{cluster=\"$cluster\"}, namespace)", + "hide": 0, + "includeAll": true, + "label": null, + "multi": false, + "name": "namespace", + "options": [ + + ], + "query": "label_values(container_network_receive_packets_total{cluster=\"$cluster\"}, namespace)", + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "auto": false, + "auto_count": 30, + "auto_min": "10s", + "current": { + "text": "5m", + "value": "5m" + }, + "datasource": "$datasource", + "hide": 0, + "includeAll": false, + "label": null, + "multi": false, + "name": "resolution", + "options": [ + { + "selected": false, + "text": "30s", + "value": "30s" + }, + { + "selected": true, + "text": "5m", + "value": "5m" + }, + { + "selected": false, + "text": "1h", + "value": "1h" + } + ], + "query": "30s,5m,1h", + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "interval", + "useTags": false + }, + { + "allValue": null, + "auto": false, + "auto_count": 30, + "auto_min": "10s", + "current": { + "text": "5m", + "value": "5m" + }, + "datasource": "$datasource", + "hide": 2, + "includeAll": false, + "label": null, + "multi": false, + "name": "interval", + "options": [ + { + "selected": true, + "text": "4h", + "value": "4h" + } + ], + "query": "4h", + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "interval", + "useTags": false + } + ] + }, + "time": { + "from": "now-1h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "5s", + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ], + "time_options": [ + "5m", + "15m", + "1h", + "6h", + "12h", + "24h", + "2d", + "7d", + "30d" + ] + }, + "timezone": "UTC", + "title": "Kubernetes / Networking / Namespace (Pods)", + "uid": "8b7a8b326d7a6f1f04244066368c67af", + "version": 0 + } + namespace-by-workload.json: |- + { + "__inputs": [ + + ], + "__requires": [ + + ], + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": "-- Grafana --", + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "type": "dashboard" + } + ] + }, + "editable": true, + "gnetId": null, + "graphTooltip": 0, + "hideControls": false, + "id": null, + "links": [ + + ], + "panels": [ + { + "collapse": false, + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 0 + }, + "id": 2, + "panels": [ + + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Current Bandwidth", + "titleSize": "h6", + "type": "row" + }, + { + "aliasColors": { + + }, + "bars": true, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 1 + }, + "id": 3, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": false, + "linewidth": 1, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "null", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 24, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\",namespace=\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{ workload }}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Current Rate of Bytes Received", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "series", + "name": null, + "show": false, + "values": [ + "current" + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": true, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 1 + }, + "id": 4, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": false, + "linewidth": 1, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "null", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 24, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\",namespace=\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{ workload }}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Current Rate of Bytes Transmitted", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "series", + "name": null, + "show": false, + "values": [ + "current" + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "columns": [ + { + "text": "Time", + "value": "Time" + }, + { + "text": "Value #A", + "value": "Value #A" + }, + { + "text": "Value #B", + "value": "Value #B" + }, + { + "text": "Value #C", + "value": "Value #C" + }, + { + "text": "Value #D", + "value": "Value #D" + }, + { + "text": "Value #E", + "value": "Value #E" + }, + { + "text": "Value #F", + "value": "Value #F" + }, + { + "text": "Value #G", + "value": "Value #G" + }, + { + "text": "Value #H", + "value": "Value #H" + }, + { + "text": "workload", + "value": "workload" + } + ], + "datasource": "$datasource", + "fill": 1, + "fontSize": "90%", + "gridPos": { + "h": 9, + "w": 24, + "x": 0, + "y": 10 + }, + "id": 5, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "null as zero", + "renderer": "flot", + "scroll": true, + "showHeader": true, + "sort": { + "col": 0, + "desc": false + }, + "spaceLength": 10, + "span": 24, + "styles": [ + { + "alias": "Time", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Time", + "thresholds": [ + + ], + "type": "hidden", + "unit": "short" + }, + { + "alias": "Current Bandwidth Received", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #A", + "thresholds": [ + + ], + "type": "number", + "unit": "Bps" + }, + { + "alias": "Current Bandwidth Transmitted", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #B", + "thresholds": [ + + ], + "type": "number", + "unit": "Bps" + }, + { + "alias": "Average Bandwidth Received", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #C", + "thresholds": [ + + ], + "type": "number", + "unit": "Bps" + }, + { + "alias": "Average Bandwidth Transmitted", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #D", + "thresholds": [ + + ], + "type": "number", + "unit": "Bps" + }, + { + "alias": "Rate of Received Packets", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #E", + "thresholds": [ + + ], + "type": "number", + "unit": "pps" + }, + { + "alias": "Rate of Transmitted Packets", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #F", + "thresholds": [ + + ], + "type": "number", + "unit": "pps" + }, + { + "alias": "Rate of Received Packets Dropped", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #G", + "thresholds": [ + + ], + "type": "number", + "unit": "pps" + }, + { + "alias": "Rate of Transmitted Packets Dropped", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #H", + "thresholds": [ + + ], + "type": "number", + "unit": "pps" + }, + { + "alias": "Workload", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": true, + "linkTooltip": "Drill down", + "linkUrl": "d/728bf77cc1166d2f3133bf25846876cc/kubernetes-networking-workload?orgId=1&refresh=30s&var-namespace=$namespace&var-type=$type&var-workload=$__cell", + "pattern": "workload", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + } + ], + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\",namespace=\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "A", + "step": 10 + }, + { + "expr": "sort_desc(sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\",namespace=\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "B", + "step": 10 + }, + { + "expr": "sort_desc(avg(irate(container_network_receive_bytes_total{cluster=\"$cluster\",namespace=\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "C", + "step": 10 + }, + { + "expr": "sort_desc(avg(irate(container_network_transmit_bytes_total{cluster=\"$cluster\",namespace=\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "D", + "step": 10 + }, + { + "expr": "sort_desc(sum(irate(container_network_receive_packets_total{cluster=\"$cluster\",namespace=\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "E", + "step": 10 + }, + { + "expr": "sort_desc(sum(irate(container_network_transmit_packets_total{cluster=\"$cluster\",namespace=\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "F", + "step": 10 + }, + { + "expr": "sort_desc(sum(irate(container_network_receive_packets_dropped_total{cluster=\"$cluster\",namespace=\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "G", + "step": 10 + }, + { + "expr": "sort_desc(sum(irate(container_network_transmit_packets_dropped_total{cluster=\"$cluster\",namespace=\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "H", + "step": 10 + } + ], + "timeFrom": null, + "timeShift": null, + "title": "Current Status", + "type": "table" + }, + { + "collapse": true, + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 19 + }, + "id": 6, + "panels": [ + { + "aliasColors": { + + }, + "bars": true, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 20 + }, + "id": 7, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": false, + "linewidth": 1, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "null", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 24, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(avg(irate(container_network_receive_bytes_total{cluster=\"$cluster\",namespace=\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{ workload }}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Average Rate of Bytes Received", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "series", + "name": null, + "show": false, + "values": [ + "current" + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": true, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 20 + }, + "id": 8, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": false, + "linewidth": 1, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "null", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 24, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(avg(irate(container_network_transmit_bytes_total{cluster=\"$cluster\",namespace=\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{ workload }}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Average Rate of Bytes Transmitted", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "series", + "name": null, + "show": false, + "values": [ + "current" + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Average Bandwidth", + "titleSize": "h6", + "type": "row" + }, + { + "collapse": false, + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 29 + }, + "id": 9, + "panels": [ + + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Bandwidth HIstory", + "titleSize": "h6", + "type": "row" + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 38 + }, + "id": 10, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\",namespace=\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{workload}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Receive Bandwidth", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 38 + }, + "id": 11, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\",namespace=\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{workload}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Transmit Bandwidth", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "collapse": true, + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 39 + }, + "id": 12, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 40 + }, + "id": 13, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_receive_packets_total{cluster=\"$cluster\",namespace=\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{workload}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of Received Packets", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 40 + }, + "id": 14, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_transmit_packets_total{cluster=\"$cluster\",namespace=\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{workload}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of Transmitted Packets", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Packets", + "titleSize": "h6", + "type": "row" + }, + { + "collapse": true, + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 40 + }, + "id": 15, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 41 + }, + "id": 16, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_receive_packets_dropped_total{cluster=\"$cluster\",namespace=\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{workload}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of Received Packets Dropped", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 41 + }, + "id": 17, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_transmit_packets_dropped_total{cluster=\"$cluster\",namespace=\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{workload}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of Transmitted Packets Dropped", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Errors", + "titleSize": "h6", + "type": "row" + } + ], + "refresh": "10s", + "rows": [ + + ], + "schemaVersion": 18, + "style": "dark", + "tags": [ + "kubernetes-mixin" + ], + "templating": { + "list": [ + { + "current": { + "text": "default", + "value": "default" + }, + "hide": 0, + "label": "Data Source", + "name": "datasource", + "options": [ + + ], + "query": "prometheus", + "refresh": 1, + "regex": "", + "type": "datasource" + }, + { + "allValue": null, + "current": { + + }, + "datasource": "$datasource", + "hide": 2, + "includeAll": false, + "label": null, + "multi": false, + "name": "cluster", + "options": [ + + ], + "query": "label_values(up{job=\"kubernetes-cadvisor\"}, cluster)", + "refresh": 2, + "regex": "", + "sort": 0, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "auto": false, + "auto_count": 30, + "auto_min": "10s", + "current": { + "text": "kube-system", + "value": "kube-system" + }, + "datasource": "$datasource", + "definition": "label_values(container_network_receive_packets_total{cluster=\"$cluster\"}, namespace)", + "hide": 0, + "includeAll": false, + "label": null, + "multi": false, + "name": "namespace", + "options": [ + + ], + "query": "label_values(container_network_receive_packets_total{cluster=\"$cluster\"}, namespace)", + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "auto": false, + "auto_count": 30, + "auto_min": "10s", + "current": { + "text": "deployment", + "value": "deployment" + }, + "datasource": "$datasource", + "definition": "label_values(namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=\"$namespace\", workload=~\".+\"}, workload_type)", + "hide": 0, + "includeAll": false, + "label": null, + "multi": false, + "name": "type", + "options": [ + + ], + "query": "label_values(namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=\"$namespace\", workload=~\".+\"}, workload_type)", + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 0, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "auto": false, + "auto_count": 30, + "auto_min": "10s", + "current": { + "text": "5m", + "value": "5m" + }, + "datasource": "$datasource", + "hide": 0, + "includeAll": false, + "label": null, + "multi": false, + "name": "resolution", + "options": [ + { + "selected": false, + "text": "30s", + "value": "30s" + }, + { + "selected": true, + "text": "5m", + "value": "5m" + }, + { + "selected": false, + "text": "1h", + "value": "1h" + } + ], + "query": "30s,5m,1h", + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "interval", + "useTags": false + }, + { + "allValue": null, + "auto": false, + "auto_count": 30, + "auto_min": "10s", + "current": { + "text": "5m", + "value": "5m" + }, + "datasource": "$datasource", + "hide": 2, + "includeAll": false, + "label": null, + "multi": false, + "name": "interval", + "options": [ + { + "selected": true, + "text": "4h", + "value": "4h" + } + ], + "query": "4h", + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "interval", + "useTags": false + } + ] + }, + "time": { + "from": "now-1h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "5s", + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ], + "time_options": [ + "5m", + "15m", + "1h", + "6h", + "12h", + "24h", + "2d", + "7d", + "30d" + ] + }, + "timezone": "UTC", + "title": "Kubernetes / Networking / Namespace (Workload)", + "uid": "bbb2a765a623ae38130206c7d94a160f", + "version": 0 + } + pod-total.json: |- + { + "__inputs": [ + + ], + "__requires": [ + + ], + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": "-- Grafana --", + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "type": "dashboard" + } + ] + }, + "editable": true, + "gnetId": null, + "graphTooltip": 0, + "hideControls": false, + "id": null, + "links": [ + + ], + "panels": [ + { + "collapse": false, + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 0 + }, + "id": 2, + "panels": [ + + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Current Bandwidth", + "titleSize": "h6", + "type": "row" + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": [ + "#299c46", + "rgba(237, 129, 40, 0.89)", + "#d44a3a" + ], + "datasource": "$datasource", + "decimals": 0, + "format": "time_series", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 1 + }, + "height": 9, + "id": 3, + "interval": null, + "links": [ + + ], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "minSpan": 12, + "nullPointMode": "connected", + "nullText": null, + "options": { + "fieldOptions": { + "calcs": [ + "last" + ], + "defaults": { + "max": 10000000000, + "min": 0, + "title": "$namespace: $pod", + "unit": "Bps" + }, + "mappings": [ + + ], + "override": { + + }, + "thresholds": [ + { + "color": "dark-green", + "index": 0, + "value": null + }, + { + "color": "dark-yellow", + "index": 1, + "value": 5000000000 + }, + { + "color": "dark-red", + "index": 2, + "value": 7000000000 + } + ], + "values": false + } + }, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 12, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\",namespace=~\"$namespace\", pod=~\"$pod\"}[$interval:$resolution]))", + "format": "time_series", + "instant": null, + "intervalFactor": 1, + "legendFormat": "", + "refId": "A" + } + ], + "thresholds": "", + "timeFrom": null, + "timeShift": null, + "title": "Current Rate of Bytes Received", + "type": "gauge", + "valueFontSize": "80%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": [ + "#299c46", + "rgba(237, 129, 40, 0.89)", + "#d44a3a" + ], + "datasource": "$datasource", + "decimals": 0, + "format": "time_series", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 1 + }, + "height": 9, + "id": 4, + "interval": null, + "links": [ + + ], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "minSpan": 12, + "nullPointMode": "connected", + "nullText": null, + "options": { + "fieldOptions": { + "calcs": [ + "last" + ], + "defaults": { + "max": 10000000000, + "min": 0, + "title": "$namespace: $pod", + "unit": "Bps" + }, + "mappings": [ + + ], + "override": { + + }, + "thresholds": [ + { + "color": "dark-green", + "index": 0, + "value": null + }, + { + "color": "dark-yellow", + "index": 1, + "value": 5000000000 + }, + { + "color": "dark-red", + "index": 2, + "value": 7000000000 + } + ], + "values": false + } + }, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 12, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\",namespace=~\"$namespace\", pod=~\"$pod\"}[$interval:$resolution]))", + "format": "time_series", + "instant": null, + "intervalFactor": 1, + "legendFormat": "", + "refId": "A" + } + ], + "thresholds": "", + "timeFrom": null, + "timeShift": null, + "title": "Current Rate of Bytes Transmitted", + "type": "gauge", + "valueFontSize": "80%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + }, + { + "collapse": false, + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 10 + }, + "id": 5, + "panels": [ + + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Bandwidth", + "titleSize": "h6", + "type": "row" + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 11 + }, + "id": 6, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\",namespace=~\"$namespace\", pod=~\"$pod\"}[$interval:$resolution])) by (pod)", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{pod}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Receive Bandwidth", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 11 + }, + "id": 7, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\",namespace=~\"$namespace\", pod=~\"$pod\"}[$interval:$resolution])) by (pod)", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{pod}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Transmit Bandwidth", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "collapse": true, + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 20 + }, + "id": 8, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 10, + "w": 12, + "x": 0, + "y": 21 + }, + "id": 9, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum(irate(container_network_receive_packets_total{cluster=\"$cluster\",namespace=~\"$namespace\", pod=~\"$pod\"}[$interval:$resolution])) by (pod)", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{pod}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of Received Packets", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 10, + "w": 12, + "x": 12, + "y": 21 + }, + "id": 10, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum(irate(container_network_transmit_packets_total{cluster=\"$cluster\",namespace=~\"$namespace\", pod=~\"$pod\"}[$interval:$resolution])) by (pod)", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{pod}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of Transmitted Packets", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Packets", + "titleSize": "h6", + "type": "row" + }, + { + "collapse": true, + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 21 + }, + "id": 11, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 10, + "w": 12, + "x": 0, + "y": 32 + }, + "id": 12, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum(irate(container_network_receive_packets_dropped_total{cluster=\"$cluster\",namespace=~\"$namespace\", pod=~\"$pod\"}[$interval:$resolution])) by (pod)", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{pod}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of Received Packets Dropped", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 10, + "w": 12, + "x": 12, + "y": 32 + }, + "id": 13, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum(irate(container_network_transmit_packets_dropped_total{cluster=\"$cluster\",namespace=~\"$namespace\", pod=~\"$pod\"}[$interval:$resolution])) by (pod)", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{pod}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of Transmitted Packets Dropped", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Errors", + "titleSize": "h6", + "type": "row" + } + ], + "refresh": "10s", + "rows": [ + + ], + "schemaVersion": 18, + "style": "dark", + "tags": [ + "kubernetes-mixin" + ], + "templating": { + "list": [ + { + "current": { + "text": "default", + "value": "default" + }, + "hide": 0, + "label": "Data Source", + "name": "datasource", + "options": [ + + ], + "query": "prometheus", + "refresh": 1, + "regex": "", + "type": "datasource" + }, + { + "allValue": null, + "current": { + + }, + "datasource": "$datasource", + "hide": 2, + "includeAll": false, + "label": null, + "multi": false, + "name": "cluster", + "options": [ + + ], + "query": "label_values(up{job=\"kubernetes-cadvisor\"}, cluster)", + "refresh": 2, + "regex": "", + "sort": 0, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": ".+", + "auto": false, + "auto_count": 30, + "auto_min": "10s", + "current": { + "text": "kube-system", + "value": "kube-system" + }, + "datasource": "$datasource", + "definition": "label_values(container_network_receive_packets_total{cluster=\"$cluster\"}, namespace)", + "hide": 0, + "includeAll": true, + "label": null, + "multi": false, + "name": "namespace", + "options": [ + + ], + "query": "label_values(container_network_receive_packets_total{cluster=\"$cluster\"}, namespace)", + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": ".+", + "auto": false, + "auto_count": 30, + "auto_min": "10s", + "current": { + "text": "", + "value": "" + }, + "datasource": "$datasource", + "definition": "label_values(container_network_receive_packets_total{cluster=\"$cluster\",namespace=~\"$namespace\"}, pod)", + "hide": 0, + "includeAll": false, + "label": null, + "multi": false, + "name": "pod", + "options": [ + + ], + "query": "label_values(container_network_receive_packets_total{cluster=\"$cluster\",namespace=~\"$namespace\"}, pod)", + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "auto": false, + "auto_count": 30, + "auto_min": "10s", + "current": { + "text": "5m", + "value": "5m" + }, + "datasource": "$datasource", + "hide": 0, + "includeAll": false, + "label": null, + "multi": false, + "name": "resolution", + "options": [ + { + "selected": false, + "text": "30s", + "value": "30s" + }, + { + "selected": true, + "text": "5m", + "value": "5m" + }, + { + "selected": false, + "text": "1h", + "value": "1h" + } + ], + "query": "30s,5m,1h", + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "interval", + "useTags": false + }, + { + "allValue": null, + "auto": false, + "auto_count": 30, + "auto_min": "10s", + "current": { + "text": "5m", + "value": "5m" + }, + "datasource": "$datasource", + "hide": 2, + "includeAll": false, + "label": null, + "multi": false, + "name": "interval", + "options": [ + { + "selected": true, + "text": "4h", + "value": "4h" + } + ], + "query": "4h", + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "interval", + "useTags": false + } + ] + }, + "time": { + "from": "now-1h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "5s", + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ], + "time_options": [ + "5m", + "15m", + "1h", + "6h", + "12h", + "24h", + "2d", + "7d", + "30d" + ] + }, + "timezone": "UTC", + "title": "Kubernetes / Networking / Pod", + "uid": "7a18067ce943a40ae25454675c19ff5c", + "version": 0 + } + workload-total.json: |- + { + "__inputs": [ + + ], + "__requires": [ + + ], + "annotations": { + "list": [ + { + "builtIn": 1, + "datasource": "-- Grafana --", + "enable": true, + "hide": true, + "iconColor": "rgba(0, 211, 255, 1)", + "name": "Annotations & Alerts", + "type": "dashboard" + } + ] + }, + "editable": true, + "gnetId": null, + "graphTooltip": 0, + "hideControls": false, + "id": null, + "links": [ + + ], + "panels": [ + { + "collapse": false, + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 0 + }, + "id": 2, + "panels": [ + + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Current Bandwidth", + "titleSize": "h6", + "type": "row" + }, + { + "aliasColors": { + + }, + "bars": true, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 1 + }, + "id": 3, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": false, + "linewidth": 1, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "null", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 24, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_receive_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{ pod }}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Current Rate of Bytes Received", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "series", + "name": null, + "show": false, + "values": [ + "current" + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": true, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 1 + }, + "id": 4, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": false, + "linewidth": 1, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "null", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 24, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_transmit_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{ pod }}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Current Rate of Bytes Transmitted", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "series", + "name": null, + "show": false, + "values": [ + "current" + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "collapse": true, + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 10 + }, + "id": 5, + "panels": [ + { + "aliasColors": { + + }, + "bars": true, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 11 + }, + "id": 6, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": false, + "linewidth": 1, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "null", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 24, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(avg(irate(container_network_receive_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{ pod }}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Average Rate of Bytes Received", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "series", + "name": null, + "show": false, + "values": [ + "current" + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": true, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 11 + }, + "id": 7, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "sort": "current", + "sortDesc": true, + "total": false, + "values": true + }, + "lines": false, + "linewidth": 1, + "links": [ + + ], + "minSpan": 24, + "nullPointMode": "null", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 24, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(avg(irate(container_network_transmit_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{ pod }}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Average Rate of Bytes Transmitted", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "series", + "name": null, + "show": false, + "values": [ + "current" + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Average Bandwidth", + "titleSize": "h6", + "type": "row" + }, + { + "collapse": false, + "collapsed": false, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 11 + }, + "id": 8, + "panels": [ + + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Bandwidth HIstory", + "titleSize": "h6", + "type": "row" + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 12 + }, + "id": 9, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_receive_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{pod}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Receive Bandwidth", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 12 + }, + "id": 10, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_transmit_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{pod}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Transmit Bandwidth", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "collapse": true, + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 21 + }, + "id": 11, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 22 + }, + "id": 12, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_receive_packets_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{pod}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of Received Packets", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 22 + }, + "id": 13, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_transmit_packets_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{pod}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of Transmitted Packets", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Packets", + "titleSize": "h6", + "type": "row" + }, + { + "collapse": true, + "collapsed": true, + "gridPos": { + "h": 1, + "w": 24, + "x": 0, + "y": 22 + }, + "id": 14, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 0, + "y": 23 + }, + "id": 15, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_receive_packets_dropped_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{pod}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of Received Packets Dropped", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 2, + "fillGradient": 0, + "gridPos": { + "h": 9, + "w": 12, + "x": 12, + "y": 23 + }, + "id": 16, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 2, + "links": [ + + ], + "minSpan": 12, + "nullPointMode": "connected", + "paceLength": 10, + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sort_desc(sum(irate(container_network_transmit_packets_dropped_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\",namespace=~\"$namespace\"}[$interval:$resolution])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "format": "time_series", + "intervalFactor": 1, + "legendFormat": "{{pod}}", + "refId": "A", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate of Transmitted Packets Dropped", + "tooltip": { + "shared": true, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Errors", + "titleSize": "h6", + "type": "row" + } + ], + "refresh": "10s", + "rows": [ + + ], + "schemaVersion": 18, + "style": "dark", + "tags": [ + "kubernetes-mixin" + ], + "templating": { + "list": [ + { + "current": { + "text": "default", + "value": "default" + }, + "hide": 0, + "label": "Data Source", + "name": "datasource", + "options": [ + + ], + "query": "prometheus", + "refresh": 1, + "regex": "", + "type": "datasource" + }, + { + "allValue": null, + "current": { + + }, + "datasource": "$datasource", + "hide": 2, + "includeAll": false, + "label": null, + "multi": false, + "name": "cluster", + "options": [ + + ], + "query": "label_values(kube_pod_info{job=\"kube-state-metrics\"}, cluster)", + "refresh": 2, + "regex": "", + "sort": 0, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": ".+", + "auto": false, + "auto_count": 30, + "auto_min": "10s", + "current": { + "text": "kube-system", + "value": "kube-system" + }, + "datasource": "$datasource", + "definition": "label_values(container_network_receive_packets_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\"}, namespace)", + "hide": 0, + "includeAll": true, + "label": null, + "multi": false, + "name": "namespace", + "options": [ + + ], + "query": "label_values(container_network_receive_packets_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\"}, namespace)", + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "auto": false, + "auto_count": 30, + "auto_min": "10s", + "current": { + "text": "", + "value": "" + }, + "datasource": "$datasource", + "definition": "label_values(namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=~\"$namespace\"}, workload)", + "hide": 0, + "includeAll": false, + "label": null, + "multi": false, + "name": "workload", + "options": [ + + ], + "query": "label_values(namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=~\"$namespace\"}, workload)", + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "auto": false, + "auto_count": 30, + "auto_min": "10s", + "current": { + "text": "deployment", + "value": "deployment" + }, + "datasource": "$datasource", + "definition": "label_values(namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=~\"$namespace\", workload=~\"$workload\"}, workload_type)", + "hide": 0, + "includeAll": false, + "label": null, + "multi": false, + "name": "type", + "options": [ + + ], + "query": "label_values(namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\",namespace=~\"$namespace\", workload=~\"$workload\"}, workload_type)", + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 0, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "auto": false, + "auto_count": 30, + "auto_min": "10s", + "current": { + "text": "5m", + "value": "5m" + }, + "datasource": "$datasource", + "hide": 0, + "includeAll": false, + "label": null, + "multi": false, + "name": "resolution", + "options": [ + { + "selected": false, + "text": "30s", + "value": "30s" + }, + { + "selected": true, + "text": "5m", + "value": "5m" + }, + { + "selected": false, + "text": "1h", + "value": "1h" + } + ], + "query": "30s,5m,1h", + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "interval", + "useTags": false + }, + { + "allValue": null, + "auto": false, + "auto_count": 30, + "auto_min": "10s", + "current": { + "text": "5m", + "value": "5m" + }, + "datasource": "$datasource", + "hide": 2, + "includeAll": false, + "label": null, + "multi": false, + "name": "interval", + "options": [ + { + "selected": true, + "text": "4h", + "value": "4h" + } + ], + "query": "4h", + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 1, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "interval", + "useTags": false + } + ] + }, + "time": { + "from": "now-1h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "5s", + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ], + "time_options": [ + "5m", + "15m", + "1h", + "6h", + "12h", + "24h", + "2d", + "7d", + "30d" + ] + }, + "timezone": "UTC", + "title": "Kubernetes / Networking / Workload", + "uid": "728bf77cc1166d2f3133bf25846876cc", + "version": 0 + } +kind: ConfigMap +metadata: + name: grafana-dashboards-k8s-network + namespace: monitoring diff --git a/addons/grafana/dashboards-k8s-nodes.yaml b/addons/grafana/dashboards-k8s-nodes.yaml index 1f34a098..2f0ef0d3 100644 --- a/addons/grafana/dashboards-k8s-nodes.yaml +++ b/addons/grafana/dashboards-k8s-nodes.yaml @@ -20,2385 +20,2106 @@ data: "id": null, "links": [ + ], + "panels": [ + { + "datasource": "$datasource", + "fieldConfig": { + "defaults": { + "links": [ + + ], + "mappings": [ + + ], + "thresholds": { + "mode": "absolute", + "steps": [ + + ] + }, + "unit": "none" + } + }, + "gridPos": { + "h": 7, + "w": 4, + "x": 0, + "y": 0 + }, + "id": 2, + "links": [ + + ], + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": [ + "lastNotNull" + ], + "fields": "", + "values": false + }, + "textMode": "auto" + }, + "pluginVersion": "7", + "targets": [ + { + "expr": "sum(kubelet_node_name{cluster=\"$cluster\", job=\"kubelet\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "", + "refId": "A" + } + ], + "title": "Running Kubelets", + "transparent": false, + "type": "stat" + }, + { + "datasource": "$datasource", + "fieldConfig": { + "defaults": { + "links": [ + + ], + "mappings": [ + + ], + "thresholds": { + "mode": "absolute", + "steps": [ + + ] + }, + "unit": "none" + } + }, + "gridPos": { + "h": 7, + "w": 4, + "x": 4, + "y": 0 + }, + "id": 3, + "links": [ + + ], + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": [ + "lastNotNull" + ], + "fields": "", + "values": false + }, + "textMode": "auto" + }, + "pluginVersion": "7", + "targets": [ + { + "expr": "sum(kubelet_running_pods{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"}) OR sum(kubelet_running_pod_count{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}}", + "refId": "A" + } + ], + "title": "Running Pods", + "transparent": false, + "type": "stat" + }, + { + "datasource": "$datasource", + "fieldConfig": { + "defaults": { + "links": [ + + ], + "mappings": [ + + ], + "thresholds": { + "mode": "absolute", + "steps": [ + + ] + }, + "unit": "none" + } + }, + "gridPos": { + "h": 7, + "w": 4, + "x": 8, + "y": 0 + }, + "id": 4, + "links": [ + + ], + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": [ + "lastNotNull" + ], + "fields": "", + "values": false + }, + "textMode": "auto" + }, + "pluginVersion": "7", + "targets": [ + { + "expr": "sum(kubelet_running_containers{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"}) OR sum(kubelet_running_container_count{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}}", + "refId": "A" + } + ], + "title": "Running Containers", + "transparent": false, + "type": "stat" + }, + { + "datasource": "$datasource", + "fieldConfig": { + "defaults": { + "links": [ + + ], + "mappings": [ + + ], + "thresholds": { + "mode": "absolute", + "steps": [ + + ] + }, + "unit": "none" + } + }, + "gridPos": { + "h": 7, + "w": 4, + "x": 12, + "y": 0 + }, + "id": 5, + "links": [ + + ], + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": [ + "lastNotNull" + ], + "fields": "", + "values": false + }, + "textMode": "auto" + }, + "pluginVersion": "7", + "targets": [ + { + "expr": "sum(volume_manager_total_volumes{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\", state=\"actual_state_of_world\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}}", + "refId": "A" + } + ], + "title": "Actual Volume Count", + "transparent": false, + "type": "stat" + }, + { + "datasource": "$datasource", + "fieldConfig": { + "defaults": { + "links": [ + + ], + "mappings": [ + + ], + "thresholds": { + "mode": "absolute", + "steps": [ + + ] + }, + "unit": "none" + } + }, + "gridPos": { + "h": 7, + "w": 4, + "x": 16, + "y": 0 + }, + "id": 6, + "links": [ + + ], + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": [ + "lastNotNull" + ], + "fields": "", + "values": false + }, + "textMode": "auto" + }, + "pluginVersion": "7", + "targets": [ + { + "expr": "sum(volume_manager_total_volumes{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\",state=\"desired_state_of_world\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}}", + "refId": "A" + } + ], + "title": "Desired Volume Count", + "transparent": false, + "type": "stat" + }, + { + "datasource": "$datasource", + "fieldConfig": { + "defaults": { + "links": [ + + ], + "mappings": [ + + ], + "thresholds": { + "mode": "absolute", + "steps": [ + + ] + }, + "unit": "none" + } + }, + "gridPos": { + "h": 7, + "w": 4, + "x": 20, + "y": 0 + }, + "id": 7, + "links": [ + + ], + "options": { + "colorMode": "value", + "graphMode": "area", + "justifyMode": "auto", + "orientation": "auto", + "reduceOptions": { + "calcs": [ + "lastNotNull" + ], + "fields": "", + "values": false + }, + "textMode": "auto" + }, + "pluginVersion": "7", + "targets": [ + { + "expr": "sum(rate(kubelet_node_config_error{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"}[$__rate_interval]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}}", + "refId": "A" + } + ], + "title": "Config Error Count", + "transparent": false, + "type": "stat" + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 7 + }, + "id": 8, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum(rate(kubelet_runtime_operations_total{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[$__rate_interval])) by (operation_type, instance)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} {{operation_type}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Operation Rate", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 7 + }, + "id": 9, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum(rate(kubelet_runtime_operations_errors_total{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[$__rate_interval])) by (instance, operation_type)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} {{operation_type}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Operation Error Rate", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 24, + "x": 0, + "y": 14 + }, + "id": 10, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.99, sum(rate(kubelet_runtime_operations_duration_seconds_bucket{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[$__rate_interval])) by (instance, operation_type, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} {{operation_type}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Operation duration 99th quantile", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 21 + }, + "id": 11, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum(rate(kubelet_pod_start_duration_seconds_count{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[$__rate_interval])) by (instance)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} pod", + "refId": "A" + }, + { + "expr": "sum(rate(kubelet_pod_worker_duration_seconds_count{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[$__rate_interval])) by (instance)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} worker", + "refId": "B" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Pod Start Rate", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 21 + }, + "id": 12, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.99, sum(rate(kubelet_pod_start_duration_seconds_count{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[$__rate_interval])) by (instance, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} pod", + "refId": "A" + }, + { + "expr": "histogram_quantile(0.99, sum(rate(kubelet_pod_worker_duration_seconds_bucket{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[$__rate_interval])) by (instance, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} worker", + "refId": "B" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Pod Start Duration", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 28 + }, + "id": 13, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum(rate(storage_operation_duration_seconds_count{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[$__rate_interval])) by (instance, operation_name, volume_plugin)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} {{operation_name}} {{volume_plugin}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Storage Operation Rate", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 28 + }, + "id": 14, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum(rate(storage_operation_errors_total{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[$__rate_interval])) by (instance, operation_name, volume_plugin)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} {{operation_name}} {{volume_plugin}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Storage Operation Error Rate", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 24, + "x": 0, + "y": 35 + }, + "id": 15, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "hideEmpty": true, + "hideZero": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.99, sum(rate(storage_operation_duration_seconds_bucket{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"}[$__rate_interval])) by (instance, operation_name, volume_plugin, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} {{operation_name}} {{volume_plugin}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Storage Operation Duration 99th quantile", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 42 + }, + "id": 16, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum(rate(kubelet_cgroup_manager_duration_seconds_count{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"}[$__rate_interval])) by (instance, operation_type)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{operation_type}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Cgroup manager operation rate", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 42 + }, + "id": 17, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.99, sum(rate(kubelet_cgroup_manager_duration_seconds_bucket{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"}[$__rate_interval])) by (instance, operation_type, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} {{operation_type}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Cgroup manager 99th quantile", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "description": "Pod lifecycle event generator", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 12, + "x": 0, + "y": 49 + }, + "id": 18, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum(rate(kubelet_pleg_relist_duration_seconds_count{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"}[$__rate_interval])) by (instance)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "PLEG relist rate", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 12, + "x": 12, + "y": 49 + }, + "id": 19, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.99, sum(rate(kubelet_pleg_relist_interval_seconds_bucket{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[$__rate_interval])) by (instance, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "PLEG relist interval", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 24, + "x": 0, + "y": 56 + }, + "id": 20, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.99, sum(rate(kubelet_pleg_relist_duration_seconds_bucket{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[$__rate_interval])) by (instance, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "PLEG relist duration", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 24, + "x": 0, + "y": 63 + }, + "id": 21, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\",job=\"kubelet\", instance=~\"$instance\",code=~\"2..\"}[$__rate_interval]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "2xx", + "refId": "A" + }, + { + "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\",job=\"kubelet\", instance=~\"$instance\",code=~\"3..\"}[$__rate_interval]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "3xx", + "refId": "B" + }, + { + "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\",job=\"kubelet\", instance=~\"$instance\",code=~\"4..\"}[$__rate_interval]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "4xx", + "refId": "C" + }, + { + "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\",job=\"kubelet\", instance=~\"$instance\",code=~\"5..\"}[$__rate_interval]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "5xx", + "refId": "D" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "RPC Rate", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 24, + "x": 0, + "y": 70 + }, + "id": 22, + "legend": { + "alignAsTable": true, + "avg": false, + "current": true, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "sideWidth": null, + "total": false, + "values": true + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{cluster=\"$cluster\",job=\"kubelet\", instance=~\"$instance\"}[$__rate_interval])) by (instance, verb, url, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} {{verb}} {{url}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Request duration 99th quantile", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 8, + "x": 0, + "y": 77 + }, + "id": 23, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "process_resident_memory_bytes{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Memory", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "bytes", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "bytes", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 8, + "x": 8, + "y": 77 + }, + "id": 24, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "rate(process_cpu_seconds_total{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[$__rate_interval])", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "CPU usage", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "fillGradient": 0, + "gridPos": { + "h": 7, + "w": 8, + "x": 16, + "y": 77 + }, + "id": 25, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "sideWidth": null, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "go_goroutines{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Goroutines", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + } ], "refresh": "10s", "rows": [ - { - "collapse": false, - "collapsed": false, - "panels": [ - { - "cacheTimeout": null, - "colorBackground": false, - "colorValue": false, - "colors": [ - "#299c46", - "rgba(237, 129, 40, 0.89)", - "#d44a3a" - ], - "datasource": "$datasource", - "format": "none", - "gauge": { - "maxValue": 100, - "minValue": 0, - "show": false, - "thresholdLabels": false, - "thresholdMarkers": true - }, - "gridPos": { - }, - "id": 2, - "interval": null, - "links": [ - - ], - "mappingType": 1, - "mappingTypes": [ - { - "name": "value to text", - "value": 1 - }, - { - "name": "range to text", - "value": 2 - } - ], - "maxDataPoints": 100, - "nullPointMode": "connected", - "nullText": null, - "postfix": "", - "postfixFontSize": "50%", - "prefix": "", - "prefixFontSize": "50%", - "rangeMaps": [ - { - "from": "null", - "text": "N/A", - "to": "null" - } - ], - "span": 2, - "sparkline": { - "fillColor": "rgba(31, 118, 189, 0.18)", - "full": false, - "lineColor": "rgb(31, 120, 193)", - "show": false - }, - "tableColumn": "", - "targets": [ - { - "expr": "sum(up{cluster=\"$cluster\", job=\"kubelet\"})", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "", - "refId": "A" - } - ], - "thresholds": "", - "title": "Up", - "tooltip": { - "shared": false - }, - "type": "singlestat", - "valueFontSize": "80%", - "valueMaps": [ - { - "op": "=", - "text": "N/A", - "value": "null" - } - ], - "valueName": "min" - }, - { - "cacheTimeout": null, - "colorBackground": false, - "colorValue": false, - "colors": [ - "#299c46", - "rgba(237, 129, 40, 0.89)", - "#d44a3a" - ], - "datasource": "$datasource", - "format": "none", - "gauge": { - "maxValue": 100, - "minValue": 0, - "show": false, - "thresholdLabels": false, - "thresholdMarkers": true - }, - "gridPos": { - - }, - "id": 3, - "interval": null, - "links": [ - - ], - "mappingType": 1, - "mappingTypes": [ - { - "name": "value to text", - "value": 1 - }, - { - "name": "range to text", - "value": 2 - } - ], - "maxDataPoints": 100, - "nullPointMode": "connected", - "nullText": null, - "postfix": "", - "postfixFontSize": "50%", - "prefix": "", - "prefixFontSize": "50%", - "rangeMaps": [ - { - "from": "null", - "text": "N/A", - "to": "null" - } - ], - "span": 2, - "sparkline": { - "fillColor": "rgba(31, 118, 189, 0.18)", - "full": false, - "lineColor": "rgb(31, 120, 193)", - "show": false - }, - "tableColumn": "", - "targets": [ - { - "expr": "sum(kubelet_running_pods{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"}) OR sum(kubelet_running_pod_count{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"})", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}}", - "refId": "A" - } - ], - "thresholds": "", - "title": "Running Pods", - "tooltip": { - "shared": false - }, - "type": "singlestat", - "valueFontSize": "80%", - "valueMaps": [ - { - "op": "=", - "text": "N/A", - "value": "null" - } - ], - "valueName": "min" - }, - { - "cacheTimeout": null, - "colorBackground": false, - "colorValue": false, - "colors": [ - "#299c46", - "rgba(237, 129, 40, 0.89)", - "#d44a3a" - ], - "datasource": "$datasource", - "format": "none", - "gauge": { - "maxValue": 100, - "minValue": 0, - "show": false, - "thresholdLabels": false, - "thresholdMarkers": true - }, - "gridPos": { - - }, - "id": 4, - "interval": null, - "links": [ - - ], - "mappingType": 1, - "mappingTypes": [ - { - "name": "value to text", - "value": 1 - }, - { - "name": "range to text", - "value": 2 - } - ], - "maxDataPoints": 100, - "nullPointMode": "connected", - "nullText": null, - "postfix": "", - "postfixFontSize": "50%", - "prefix": "", - "prefixFontSize": "50%", - "rangeMaps": [ - { - "from": "null", - "text": "N/A", - "to": "null" - } - ], - "span": 2, - "sparkline": { - "fillColor": "rgba(31, 118, 189, 0.18)", - "full": false, - "lineColor": "rgb(31, 120, 193)", - "show": false - }, - "tableColumn": "", - "targets": [ - { - "expr": "sum(kubelet_running_containers{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"}) OR sum(kubelet_running_container_count{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"})", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}}", - "refId": "A" - } - ], - "thresholds": "", - "title": "Running Container", - "tooltip": { - "shared": false - }, - "type": "singlestat", - "valueFontSize": "80%", - "valueMaps": [ - { - "op": "=", - "text": "N/A", - "value": "null" - } - ], - "valueName": "min" - }, - { - "cacheTimeout": null, - "colorBackground": false, - "colorValue": false, - "colors": [ - "#299c46", - "rgba(237, 129, 40, 0.89)", - "#d44a3a" - ], - "datasource": "$datasource", - "format": "none", - "gauge": { - "maxValue": 100, - "minValue": 0, - "show": false, - "thresholdLabels": false, - "thresholdMarkers": true - }, - "gridPos": { - - }, - "id": 5, - "interval": null, - "links": [ - - ], - "mappingType": 1, - "mappingTypes": [ - { - "name": "value to text", - "value": 1 - }, - { - "name": "range to text", - "value": 2 - } - ], - "maxDataPoints": 100, - "nullPointMode": "connected", - "nullText": null, - "postfix": "", - "postfixFontSize": "50%", - "prefix": "", - "prefixFontSize": "50%", - "rangeMaps": [ - { - "from": "null", - "text": "N/A", - "to": "null" - } - ], - "span": 2, - "sparkline": { - "fillColor": "rgba(31, 118, 189, 0.18)", - "full": false, - "lineColor": "rgb(31, 120, 193)", - "show": false - }, - "tableColumn": "", - "targets": [ - { - "expr": "sum(volume_manager_total_volumes{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\", state=\"actual_state_of_world\"})", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}}", - "refId": "A" - } - ], - "thresholds": "", - "title": "Actual Volume Count", - "tooltip": { - "shared": false - }, - "type": "singlestat", - "valueFontSize": "80%", - "valueMaps": [ - { - "op": "=", - "text": "N/A", - "value": "null" - } - ], - "valueName": "min" - }, - { - "cacheTimeout": null, - "colorBackground": false, - "colorValue": false, - "colors": [ - "#299c46", - "rgba(237, 129, 40, 0.89)", - "#d44a3a" - ], - "datasource": "$datasource", - "format": "none", - "gauge": { - "maxValue": 100, - "minValue": 0, - "show": false, - "thresholdLabels": false, - "thresholdMarkers": true - }, - "gridPos": { - - }, - "id": 6, - "interval": null, - "links": [ - - ], - "mappingType": 1, - "mappingTypes": [ - { - "name": "value to text", - "value": 1 - }, - { - "name": "range to text", - "value": 2 - } - ], - "maxDataPoints": 100, - "nullPointMode": "connected", - "nullText": null, - "postfix": "", - "postfixFontSize": "50%", - "prefix": "", - "prefixFontSize": "50%", - "rangeMaps": [ - { - "from": "null", - "text": "N/A", - "to": "null" - } - ], - "span": 2, - "sparkline": { - "fillColor": "rgba(31, 118, 189, 0.18)", - "full": false, - "lineColor": "rgb(31, 120, 193)", - "show": false - }, - "tableColumn": "", - "targets": [ - { - "expr": "sum(volume_manager_total_volumes{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\",state=\"desired_state_of_world\"})", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}}", - "refId": "A" - } - ], - "thresholds": "", - "title": "Desired Volume Count", - "tooltip": { - "shared": false - }, - "type": "singlestat", - "valueFontSize": "80%", - "valueMaps": [ - { - "op": "=", - "text": "N/A", - "value": "null" - } - ], - "valueName": "min" - }, - { - "cacheTimeout": null, - "colorBackground": false, - "colorValue": false, - "colors": [ - "#299c46", - "rgba(237, 129, 40, 0.89)", - "#d44a3a" - ], - "datasource": "$datasource", - "format": "none", - "gauge": { - "maxValue": 100, - "minValue": 0, - "show": false, - "thresholdLabels": false, - "thresholdMarkers": true - }, - "gridPos": { - - }, - "id": 7, - "interval": null, - "links": [ - - ], - "mappingType": 1, - "mappingTypes": [ - { - "name": "value to text", - "value": 1 - }, - { - "name": "range to text", - "value": 2 - } - ], - "maxDataPoints": 100, - "nullPointMode": "connected", - "nullText": null, - "postfix": "", - "postfixFontSize": "50%", - "prefix": "", - "prefixFontSize": "50%", - "rangeMaps": [ - { - "from": "null", - "text": "N/A", - "to": "null" - } - ], - "span": 2, - "sparkline": { - "fillColor": "rgba(31, 118, 189, 0.18)", - "full": false, - "lineColor": "rgb(31, 120, 193)", - "show": false - }, - "tableColumn": "", - "targets": [ - { - "expr": "sum(rate(kubelet_node_config_error{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"}[5m]))", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}}", - "refId": "A" - } - ], - "thresholds": "", - "title": "Config Error Count", - "tooltip": { - "shared": false - }, - "type": "singlestat", - "valueFontSize": "80%", - "valueMaps": [ - { - "op": "=", - "text": "N/A", - "value": "null" - } - ], - "valueName": "min" - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Dashboard Row", - "titleSize": "h6", - "type": "row" - }, - { - "collapse": false, - "collapsed": false, - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "fillGradient": 0, - "gridPos": { - - }, - "id": 8, - "legend": { - "alignAsTable": true, - "avg": false, - "current": true, - "max": false, - "min": false, - "rightSide": true, - "show": true, - "sideWidth": null, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "sum(rate(kubelet_runtime_operations_total{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (operation_type, instance)", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}} {{operation_type}}", - "refId": "A" - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Operation Rate", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "ops", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": true - }, - { - "format": "ops", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": true - } - ] - }, - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "fillGradient": 0, - "gridPos": { - - }, - "id": 9, - "legend": { - "alignAsTable": true, - "avg": false, - "current": true, - "max": false, - "min": false, - "rightSide": true, - "show": true, - "sideWidth": null, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "sum(rate(kubelet_runtime_operations_errors_total{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, operation_type)", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}} {{operation_type}}", - "refId": "A" - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Operation Error Rate", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "ops", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - }, - { - "format": "ops", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - } - ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Dashboard Row", - "titleSize": "h6", - "type": "row" - }, - { - "collapse": false, - "collapsed": false, - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "fillGradient": 0, - "gridPos": { - - }, - "id": 10, - "legend": { - "alignAsTable": true, - "avg": false, - "current": true, - "max": false, - "min": false, - "rightSide": true, - "show": true, - "sideWidth": null, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 12, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "histogram_quantile(0.99, sum(rate(kubelet_runtime_operations_duration_seconds_bucket{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, operation_type, le))", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}} {{operation_type}}", - "refId": "A" - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Operation duration 99th quantile", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "s", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": true - }, - { - "format": "s", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": true - } - ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Dashboard Row", - "titleSize": "h6", - "type": "row" - }, - { - "collapse": false, - "collapsed": false, - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "fillGradient": 0, - "gridPos": { - - }, - "id": 11, - "legend": { - "alignAsTable": true, - "avg": false, - "current": true, - "max": false, - "min": false, - "rightSide": true, - "show": true, - "sideWidth": null, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "sum(rate(kubelet_pod_start_duration_seconds_count{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance)", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}} pod", - "refId": "A" - }, - { - "expr": "sum(rate(kubelet_pod_worker_duration_seconds_count{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance)", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}} worker", - "refId": "B" - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Pod Start Rate", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "ops", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - }, - { - "format": "ops", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - } - ] - }, - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "fillGradient": 0, - "gridPos": { - - }, - "id": 12, - "legend": { - "alignAsTable": true, - "avg": false, - "current": true, - "max": false, - "min": false, - "rightSide": true, - "show": true, - "sideWidth": null, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "histogram_quantile(0.99, sum(rate(kubelet_pod_start_duration_seconds_count{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, le))", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}} pod", - "refId": "A" - }, - { - "expr": "histogram_quantile(0.99, sum(rate(kubelet_pod_worker_duration_seconds_bucket{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, le))", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}} worker", - "refId": "B" - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Pod Start Duration", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "s", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - }, - { - "format": "s", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - } - ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Dashboard Row", - "titleSize": "h6", - "type": "row" - }, - { - "collapse": false, - "collapsed": false, - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "fillGradient": 0, - "gridPos": { - - }, - "id": 13, - "legend": { - "alignAsTable": true, - "avg": false, - "current": true, - "hideEmpty": true, - "hideZero": true, - "max": false, - "min": false, - "rightSide": true, - "show": true, - "sideWidth": null, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "sum(rate(storage_operation_duration_seconds_count{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, operation_name, volume_plugin)", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}} {{operation_name}} {{volume_plugin}}", - "refId": "A" - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Storage Operation Rate", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "ops", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - }, - { - "format": "ops", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - } - ] - }, - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "fillGradient": 0, - "gridPos": { - - }, - "id": 14, - "legend": { - "alignAsTable": true, - "avg": false, - "current": true, - "hideEmpty": true, - "hideZero": true, - "max": false, - "min": false, - "rightSide": true, - "show": true, - "sideWidth": null, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "sum(rate(storage_operation_errors_total{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, operation_name, volume_plugin)", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}} {{operation_name}} {{volume_plugin}}", - "refId": "A" - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Storage Operation Error Rate", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "ops", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - }, - { - "format": "ops", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - } - ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Dashboard Row", - "titleSize": "h6", - "type": "row" - }, - { - "collapse": false, - "collapsed": false, - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "fillGradient": 0, - "gridPos": { - - }, - "id": 15, - "legend": { - "alignAsTable": true, - "avg": false, - "current": true, - "hideEmpty": true, - "hideZero": true, - "max": false, - "min": false, - "rightSide": true, - "show": true, - "sideWidth": null, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 12, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "histogram_quantile(0.99, sum(rate(storage_operation_duration_seconds_bucket{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"}[5m])) by (instance, operation_name, volume_plugin, le))", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}} {{operation_name}} {{volume_plugin}}", - "refId": "A" - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Storage Operation Duration 99th quantile", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "s", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - }, - { - "format": "s", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - } - ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Dashboard Row", - "titleSize": "h6", - "type": "row" - }, - { - "collapse": false, - "collapsed": false, - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "fillGradient": 0, - "gridPos": { - - }, - "id": 16, - "legend": { - "alignAsTable": true, - "avg": false, - "current": true, - "max": false, - "min": false, - "rightSide": true, - "show": true, - "sideWidth": null, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "sum(rate(kubelet_cgroup_manager_duration_seconds_count{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"}[5m])) by (instance, operation_type)", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{operation_type}}", - "refId": "A" - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Cgroup manager operation rate", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "ops", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - }, - { - "format": "ops", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - } - ] - }, - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "fillGradient": 0, - "gridPos": { - - }, - "id": 17, - "legend": { - "alignAsTable": true, - "avg": false, - "current": true, - "max": false, - "min": false, - "rightSide": true, - "show": true, - "sideWidth": null, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "histogram_quantile(0.99, sum(rate(kubelet_cgroup_manager_duration_seconds_bucket{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"}[5m])) by (instance, operation_type, le))", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}} {{operation_type}}", - "refId": "A" - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Cgroup manager 99th quantile", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "s", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - }, - { - "format": "s", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - } - ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Dashboard Row", - "titleSize": "h6", - "type": "row" - }, - { - "collapse": false, - "collapsed": false, - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "description": "Pod lifecycle event generator", - "fill": 1, - "fillGradient": 0, - "gridPos": { - - }, - "id": 18, - "legend": { - "alignAsTable": true, - "avg": false, - "current": true, - "max": false, - "min": false, - "rightSide": true, - "show": true, - "sideWidth": null, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "sum(rate(kubelet_pleg_relist_duration_seconds_count{cluster=\"$cluster\", job=\"kubelet\", instance=~\"$instance\"}[5m])) by (instance)", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}}", - "refId": "A" - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "PLEG relist rate", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "ops", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - }, - { - "format": "ops", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - } - ] - }, - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "fillGradient": 0, - "gridPos": { - - }, - "id": 19, - "legend": { - "alignAsTable": true, - "avg": false, - "current": true, - "max": false, - "min": false, - "rightSide": true, - "show": true, - "sideWidth": null, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "histogram_quantile(0.99, sum(rate(kubelet_pleg_relist_interval_seconds_bucket{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, le))", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}}", - "refId": "A" - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "PLEG relist interval", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "s", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - }, - { - "format": "s", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - } - ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Dashboard Row", - "titleSize": "h6", - "type": "row" - }, - { - "collapse": false, - "collapsed": false, - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "fillGradient": 0, - "gridPos": { - - }, - "id": 20, - "legend": { - "alignAsTable": true, - "avg": false, - "current": true, - "max": false, - "min": false, - "rightSide": true, - "show": true, - "sideWidth": null, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 12, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "histogram_quantile(0.99, sum(rate(kubelet_pleg_relist_duration_seconds_bucket{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance, le))", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}}", - "refId": "A" - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "PLEG relist duration", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "s", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - }, - { - "format": "s", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - } - ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Dashboard Row", - "titleSize": "h6", - "type": "row" - }, - { - "collapse": false, - "collapsed": false, - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "fillGradient": 0, - "gridPos": { - - }, - "id": 21, - "legend": { - "alignAsTable": false, - "avg": false, - "current": false, - "max": false, - "min": false, - "rightSide": false, - "show": true, - "sideWidth": null, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 12, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\",job=\"kubelet\", instance=~\"$instance\",code=~\"2..\"}[5m]))", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "2xx", - "refId": "A" - }, - { - "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\",job=\"kubelet\", instance=~\"$instance\",code=~\"3..\"}[5m]))", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "3xx", - "refId": "B" - }, - { - "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\",job=\"kubelet\", instance=~\"$instance\",code=~\"4..\"}[5m]))", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "4xx", - "refId": "C" - }, - { - "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\",job=\"kubelet\", instance=~\"$instance\",code=~\"5..\"}[5m]))", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "5xx", - "refId": "D" - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "RPC Rate", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "ops", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - }, - { - "format": "ops", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - } - ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Dashboard Row", - "titleSize": "h6", - "type": "row" - }, - { - "collapse": false, - "collapsed": false, - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "fillGradient": 0, - "gridPos": { - - }, - "id": 22, - "legend": { - "alignAsTable": true, - "avg": false, - "current": true, - "max": false, - "min": false, - "rightSide": true, - "show": true, - "sideWidth": null, - "total": false, - "values": true - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 12, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{cluster=\"$cluster\",job=\"kubelet\", instance=~\"$instance\"}[5m])) by (instance, verb, url, le))", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}} {{verb}} {{url}}", - "refId": "A" - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Request duration 99th quantile", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "s", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - }, - { - "format": "s", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - } - ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Dashboard Row", - "titleSize": "h6", - "type": "row" - }, - { - "collapse": false, - "collapsed": false, - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "fillGradient": 0, - "gridPos": { - - }, - "id": 23, - "legend": { - "alignAsTable": false, - "avg": false, - "current": false, - "max": false, - "min": false, - "rightSide": false, - "show": true, - "sideWidth": null, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 4, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "process_resident_memory_bytes{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}}", - "refId": "A" - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Memory", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "bytes", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": true - }, - { - "format": "bytes", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": true - } - ] - }, - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "fillGradient": 0, - "gridPos": { - - }, - "id": 24, - "legend": { - "alignAsTable": false, - "avg": false, - "current": false, - "max": false, - "min": false, - "rightSide": false, - "show": true, - "sideWidth": null, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 4, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "rate(process_cpu_seconds_total{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}[5m])", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}}", - "refId": "A" - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "CPU usage", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - }, - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - } - ] - }, - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "fillGradient": 0, - "gridPos": { - - }, - "id": 25, - "legend": { - "alignAsTable": false, - "avg": false, - "current": false, - "max": false, - "min": false, - "rightSide": false, - "show": true, - "sideWidth": null, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 4, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "go_goroutines{cluster=\"$cluster\",job=\"kubelet\",instance=~\"$instance\"}", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{instance}}", - "refId": "A" - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Goroutines", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": true - }, - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": true - } - ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Dashboard Row", - "titleSize": "h6", - "type": "row" - } ], "schemaVersion": 14, "style": "dark", @@ -2413,7 +2134,7 @@ data: "value": "default" }, "hide": 0, - "label": null, + "label": "Data Source", "name": "datasource", "options": [ @@ -2437,7 +2158,7 @@ data: "options": [ ], - "query": "label_values(kube_pod_info, cluster)", + "query": "label_values(up{job=\"kubelet\"}, cluster)", "refresh": 2, "regex": "", "sort": 1, @@ -2457,13 +2178,13 @@ data: "datasource": "$datasource", "hide": 0, "includeAll": true, - "label": null, + "label": "instance", "multi": false, "name": "instance", "options": [ ], - "query": "label_values(kubelet_runtime_operations_total{cluster=\"$cluster\", job=\"kubelet\"}, instance)", + "query": "label_values(up{job=\"kubelet\",cluster=\"$cluster\"}, instance)", "refresh": 2, "regex": "", "sort": 1, @@ -2560,7 +2281,11 @@ data: }, "id": 2, - "interval": null, + "interval": "1m", + "legend": { + "alignAsTable": true, + "rightSide": true + }, "links": [ ], @@ -2599,7 +2324,7 @@ data: "tableColumn": "", "targets": [ { - "expr": "sum(up{job=\"kube-proxy\"})", + "expr": "sum(up{cluster=\"$cluster\", job=\"kube-proxy\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "", @@ -2636,13 +2361,14 @@ data: }, "id": 3, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -2668,7 +2394,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(rate(kubeproxy_sync_proxy_rules_duration_seconds_count{job=\"kube-proxy\", instance=~\"$instance\"}[5m]))", + "expr": "sum(rate(kubeproxy_sync_proxy_rules_duration_seconds_count{cluster=\"$cluster\", job=\"kube-proxy\", instance=~\"$instance\"}[$__rate_interval]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "rate", @@ -2729,6 +2455,7 @@ data: }, "id": 4, + "interval": "1m", "legend": { "alignAsTable": true, "avg": false, @@ -2761,7 +2488,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "histogram_quantile(0.99,rate(kubeproxy_sync_proxy_rules_duration_seconds_bucket{job=\"kube-proxy\", instance=~\"$instance\"}[5m]))", + "expr": "histogram_quantile(0.99,rate(kubeproxy_sync_proxy_rules_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-proxy\", instance=~\"$instance\"}[$__rate_interval]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{instance}}", @@ -2835,13 +2562,14 @@ data: }, "id": 5, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -2867,7 +2595,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(rate(kubeproxy_network_programming_duration_seconds_count{job=\"kube-proxy\", instance=~\"$instance\"}[5m]))", + "expr": "sum(rate(kubeproxy_network_programming_duration_seconds_count{cluster=\"$cluster\", job=\"kube-proxy\", instance=~\"$instance\"}[$__rate_interval]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "rate", @@ -2928,6 +2656,7 @@ data: }, "id": 6, + "interval": "1m", "legend": { "alignAsTable": true, "avg": false, @@ -2960,7 +2689,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "histogram_quantile(0.99, sum(rate(kubeproxy_network_programming_duration_seconds_bucket{job=\"kube-proxy\", instance=~\"$instance\"}[5m])) by (instance, le))", + "expr": "histogram_quantile(0.99, sum(rate(kubeproxy_network_programming_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-proxy\", instance=~\"$instance\"}[$__rate_interval])) by (instance, le))", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{instance}}", @@ -3034,13 +2763,14 @@ data: }, "id": 7, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -3066,28 +2796,28 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(rate(rest_client_requests_total{job=\"kube-proxy\", instance=~\"$instance\",code=~\"2..\"}[5m]))", + "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\", job=\"kube-proxy\", instance=~\"$instance\",code=~\"2..\"}[$__rate_interval]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "2xx", "refId": "A" }, { - "expr": "sum(rate(rest_client_requests_total{job=\"kube-proxy\", instance=~\"$instance\",code=~\"3..\"}[5m]))", + "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\", job=\"kube-proxy\", instance=~\"$instance\",code=~\"3..\"}[$__rate_interval]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "3xx", "refId": "B" }, { - "expr": "sum(rate(rest_client_requests_total{job=\"kube-proxy\", instance=~\"$instance\",code=~\"4..\"}[5m]))", + "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\", job=\"kube-proxy\", instance=~\"$instance\",code=~\"4..\"}[$__rate_interval]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "4xx", "refId": "C" }, { - "expr": "sum(rate(rest_client_requests_total{job=\"kube-proxy\", instance=~\"$instance\",code=~\"5..\"}[5m]))", + "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\", job=\"kube-proxy\", instance=~\"$instance\",code=~\"5..\"}[$__rate_interval]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "5xx", @@ -3148,13 +2878,14 @@ data: }, "id": 8, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -3180,7 +2911,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{job=\"kube-proxy\",instance=~\"$instance\",verb=\"POST\"}[5m])) by (verb, url, le))", + "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-proxy\",instance=~\"$instance\",verb=\"POST\"}[$__rate_interval])) by (verb, url, le))", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{verb}} {{url}}", @@ -3254,6 +2985,7 @@ data: }, "id": 9, + "interval": "1m", "legend": { "alignAsTable": true, "avg": false, @@ -3286,7 +3018,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{job=\"kube-proxy\", instance=~\"$instance\", verb=\"GET\"}[5m])) by (verb, url, le))", + "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-proxy\", instance=~\"$instance\", verb=\"GET\"}[$__rate_interval])) by (verb, url, le))", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{verb}} {{url}}", @@ -3360,13 +3092,14 @@ data: }, "id": 10, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -3392,7 +3125,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "process_resident_memory_bytes{job=\"kube-proxy\",instance=~\"$instance\"}", + "expr": "process_resident_memory_bytes{cluster=\"$cluster\", job=\"kube-proxy\",instance=~\"$instance\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{instance}}", @@ -3453,13 +3186,14 @@ data: }, "id": 11, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -3485,7 +3219,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "rate(process_cpu_seconds_total{job=\"kube-proxy\",instance=~\"$instance\"}[5m])", + "expr": "rate(process_cpu_seconds_total{cluster=\"$cluster\", job=\"kube-proxy\",instance=~\"$instance\"}[$__rate_interval])", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{instance}}", @@ -3546,13 +3280,14 @@ data: }, "id": 12, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -3578,7 +3313,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "go_goroutines{job=\"kube-proxy\",instance=~\"$instance\"}", + "expr": "go_goroutines{cluster=\"$cluster\", job=\"kube-proxy\",instance=~\"$instance\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{instance}}", @@ -3648,7 +3383,7 @@ data: "value": "default" }, "hide": 0, - "label": null, + "label": "Data Source", "name": "datasource", "options": [ @@ -3662,6 +3397,32 @@ data: "allValue": null, "current": { + }, + "datasource": "$datasource", + "hide": 2, + "includeAll": false, + "label": "cluster", + "multi": false, + "name": "cluster", + "options": [ + + ], + "query": "label_values(up{job=\"kube-proxy\"}, cluster)", + "refresh": 2, + "regex": "", + "sort": 1, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "current": { + }, "datasource": "$datasource", "hide": 0, @@ -3672,7 +3433,7 @@ data: "options": [ ], - "query": "label_values(kubeproxy_network_programming_duration_seconds_bucket{job=\"kube-proxy\"}, instance)", + "query": "label_values(up{job=\"kube-proxy\", cluster=\"$cluster\", job=\"kube-proxy\"}, instance)", "refresh": 2, "regex": "", "sort": 1, diff --git a/addons/grafana/dashboards-k8s-resources-1.yaml b/addons/grafana/dashboards-k8s-resources-1.yaml index 860de6bb..0983da68 100644 --- a/addons/grafana/dashboards-k8s-resources-1.yaml +++ b/addons/grafana/dashboards-k8s-resources-1.yaml @@ -33,10 +33,12 @@ data: "id": 1, "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -60,7 +62,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "1 - avg(rate(node_cpu_seconds_total{mode=\"idle\", cluster=\"$cluster\"}[$__rate_interval]))", + "expr": "cluster:node_cpu:ratio_rate5m{cluster=\"$cluster\"}", "format": "time_series", "instant": true, "intervalFactor": 2, @@ -73,7 +75,7 @@ data: "title": "CPU Utilisation", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "singlestat", @@ -116,11 +118,14 @@ data: "fill": 1, "format": "percentunit", "id": 2, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -144,7 +149,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable_cpu_cores{cluster=\"$cluster\"})", + "expr": "sum(namespace_cpu:kube_pod_container_resource_requests:sum{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable{job=\"kube-state-metrics\",resource=\"cpu\",cluster=\"$cluster\"})", "format": "time_series", "instant": true, "intervalFactor": 2, @@ -157,7 +162,7 @@ data: "title": "CPU Requests Commitment", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "singlestat", @@ -200,11 +205,14 @@ data: "fill": 1, "format": "percentunit", "id": 3, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -228,7 +236,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable_cpu_cores{cluster=\"$cluster\"})", + "expr": "sum(namespace_cpu:kube_pod_container_resource_limits:sum{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable{job=\"kube-state-metrics\",resource=\"cpu\",cluster=\"$cluster\"})", "format": "time_series", "instant": true, "intervalFactor": 2, @@ -241,7 +249,7 @@ data: "title": "CPU Limits Commitment", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "singlestat", @@ -284,11 +292,14 @@ data: "fill": 1, "format": "percentunit", "id": 4, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -312,7 +323,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "1 - sum(:node_memory_MemAvailable_bytes:sum{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable_memory_bytes{cluster=\"$cluster\"})", + "expr": "1 - sum(:node_memory_MemAvailable_bytes:sum{cluster=\"$cluster\"}) / sum(node_memory_MemTotal_bytes{job=\"node-exporter\",cluster=\"$cluster\"})", "format": "time_series", "instant": true, "intervalFactor": 2, @@ -325,7 +336,7 @@ data: "title": "Memory Utilisation", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "singlestat", @@ -368,11 +379,14 @@ data: "fill": 1, "format": "percentunit", "id": 5, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -396,7 +410,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable_memory_bytes{cluster=\"$cluster\"})", + "expr": "sum(namespace_memory:kube_pod_container_resource_requests:sum{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable{job=\"kube-state-metrics\",resource=\"memory\",cluster=\"$cluster\"})", "format": "time_series", "instant": true, "intervalFactor": 2, @@ -409,7 +423,7 @@ data: "title": "Memory Requests Commitment", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "singlestat", @@ -452,11 +466,14 @@ data: "fill": 1, "format": "percentunit", "id": 6, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -480,7 +497,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable_memory_bytes{cluster=\"$cluster\"})", + "expr": "sum(namespace_memory:kube_pod_container_resource_limits:sum{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable{job=\"kube-state-metrics\",resource=\"memory\",cluster=\"$cluster\"})", "format": "time_series", "instant": true, "intervalFactor": 2, @@ -493,7 +510,7 @@ data: "title": "Memory Limits Commitment", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "singlestat", @@ -547,11 +564,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 7, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -575,7 +595,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\"}) by (namespace)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{namespace}}", @@ -591,7 +611,7 @@ data: "title": "CPU Usage", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -645,11 +665,14 @@ data: "datasource": "$datasource", "fill": 1, "id": 8, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -689,7 +712,7 @@ data: "link": true, "linkTargetBlank": false, "linkTooltip": "Drill down to pods", - "linkUrl": "./d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell_1", + "linkUrl": "/d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell_1", "pattern": "Value #A", "thresholds": [ @@ -708,7 +731,7 @@ data: "link": true, "linkTargetBlank": false, "linkTooltip": "Drill down to workloads", - "linkUrl": "./d/a87fb0d919ec0ea5f6543124e16c42a5/k8s-resources-workloads-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell_1", + "linkUrl": "/d/a87fb0d919ec0ea5f6543124e16c42a5/k8s-resources-workloads-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell_1", "pattern": "Value #B", "thresholds": [ @@ -822,7 +845,7 @@ data: "link": true, "linkTargetBlank": false, "linkTooltip": "Drill down to pods", - "linkUrl": "./d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell", + "linkUrl": "/d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell", "pattern": "namespace", "thresholds": [ @@ -848,7 +871,7 @@ data: ], "targets": [ { - "expr": "sum(kube_pod_owner{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(kube_pod_owner{job=\"kube-state-metrics\", cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -866,7 +889,7 @@ data: "step": 10 }, { - "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -875,7 +898,7 @@ data: "step": 10 }, { - "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(namespace_cpu:kube_pod_container_resource_requests:sum{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -884,7 +907,7 @@ data: "step": 10 }, { - "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\"}) by (namespace) / sum(namespace_cpu:kube_pod_container_resource_requests:sum{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -893,7 +916,7 @@ data: "step": 10 }, { - "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(namespace_cpu:kube_pod_container_resource_limits:sum{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -902,7 +925,7 @@ data: "step": 10 }, { - "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\"}) by (namespace) / sum(namespace_cpu:kube_pod_container_resource_limits:sum{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -919,7 +942,7 @@ data: "title": "CPU Quota", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "transform": "table", @@ -974,11 +997,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 9, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -1002,7 +1028,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(container_memory_rss{cluster=\"$cluster\", container!=\"\"}) by (namespace)", + "expr": "sum(container_memory_rss{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", container!=\"\"}) by (namespace)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{namespace}}", @@ -1018,7 +1044,7 @@ data: "title": "Memory Usage (w/o cache)", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -1072,11 +1098,14 @@ data: "datasource": "$datasource", "fill": 1, "id": 10, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -1116,7 +1145,7 @@ data: "link": true, "linkTargetBlank": false, "linkTooltip": "Drill down to pods", - "linkUrl": "./d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell_1", + "linkUrl": "/d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell_1", "pattern": "Value #A", "thresholds": [ @@ -1135,7 +1164,7 @@ data: "link": true, "linkTargetBlank": false, "linkTooltip": "Drill down to workloads", - "linkUrl": "./d/a87fb0d919ec0ea5f6543124e16c42a5/k8s-resources-workloads-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell_1", + "linkUrl": "/d/a87fb0d919ec0ea5f6543124e16c42a5/k8s-resources-workloads-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell_1", "pattern": "Value #B", "thresholds": [ @@ -1249,7 +1278,7 @@ data: "link": true, "linkTargetBlank": false, "linkTooltip": "Drill down to pods", - "linkUrl": "./d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell", + "linkUrl": "/d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell", "pattern": "namespace", "thresholds": [ @@ -1275,7 +1304,7 @@ data: ], "targets": [ { - "expr": "sum(kube_pod_owner{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(kube_pod_owner{job=\"kube-state-metrics\", cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -1293,7 +1322,7 @@ data: "step": 10 }, { - "expr": "sum(container_memory_rss{cluster=\"$cluster\", container!=\"\"}) by (namespace)", + "expr": "sum(container_memory_rss{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", container!=\"\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -1302,7 +1331,7 @@ data: "step": 10 }, { - "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(namespace_memory:kube_pod_container_resource_requests:sum{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -1311,7 +1340,7 @@ data: "step": 10 }, { - "expr": "sum(container_memory_rss{cluster=\"$cluster\", container!=\"\"}) by (namespace) / sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(container_memory_rss{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", container!=\"\"}) by (namespace) / sum(namespace_memory:kube_pod_container_resource_requests:sum{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -1320,7 +1349,7 @@ data: "step": 10 }, { - "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(namespace_memory:kube_pod_container_resource_limits:sum{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -1329,7 +1358,7 @@ data: "step": 10 }, { - "expr": "sum(container_memory_rss{cluster=\"$cluster\", container!=\"\"}) by (namespace) / sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(container_memory_rss{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", container!=\"\"}) by (namespace) / sum(namespace_memory:kube_pod_container_resource_limits:sum{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -1346,7 +1375,7 @@ data: "title": "Requests by Namespace", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "transform": "table", @@ -1403,10 +1432,12 @@ data: "id": 11, "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -1560,7 +1591,7 @@ data: "link": true, "linkTargetBlank": false, "linkTooltip": "Drill down to pods", - "linkUrl": "./d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell", + "linkUrl": "/d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell", "pattern": "namespace", "thresholds": [ @@ -1586,7 +1617,7 @@ data: ], "targets": [ { - "expr": "sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", + "expr": "sum(irate(container_network_receive_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -1595,7 +1626,7 @@ data: "step": 10 }, { - "expr": "sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", + "expr": "sum(irate(container_network_transmit_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -1604,7 +1635,7 @@ data: "step": 10 }, { - "expr": "sum(irate(container_network_receive_packets_total{cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", + "expr": "sum(irate(container_network_receive_packets_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -1613,7 +1644,7 @@ data: "step": 10 }, { - "expr": "sum(irate(container_network_transmit_packets_total{cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", + "expr": "sum(irate(container_network_transmit_packets_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -1622,7 +1653,7 @@ data: "step": 10 }, { - "expr": "sum(irate(container_network_receive_packets_dropped_total{cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", + "expr": "sum(irate(container_network_receive_packets_dropped_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -1631,7 +1662,7 @@ data: "step": 10 }, { - "expr": "sum(irate(container_network_transmit_packets_dropped_total{cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", + "expr": "sum(irate(container_network_transmit_packets_dropped_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -1648,7 +1679,7 @@ data: "title": "Current Network Usage", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "transform": "table", @@ -1686,7 +1717,7 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Current Network Usage", "titleSize": "h6" }, { @@ -1703,11 +1734,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 12, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -1726,12 +1760,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", + "expr": "sum(irate(container_network_receive_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{namespace}}", @@ -1747,7 +1781,7 @@ data: "title": "Receive Bandwidth", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -1778,19 +1812,7 @@ data: "show": false } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Network", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ + }, { "aliasColors": { @@ -1801,11 +1823,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 13, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -1824,12 +1849,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", + "expr": "sum(irate(container_network_transmit_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{namespace}}", @@ -1845,7 +1870,7 @@ data: "title": "Transmit Bandwidth", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -1882,7 +1907,7 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Bandwidth", "titleSize": "h6" }, { @@ -1899,11 +1924,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 14, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -1922,12 +1950,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "avg(irate(container_network_receive_bytes_total{cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", + "expr": "avg(irate(container_network_receive_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{namespace}}", @@ -1943,7 +1971,7 @@ data: "title": "Average Container Bandwidth by Namespace: Received", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -1974,19 +2002,7 @@ data: "show": false } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Network", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ + }, { "aliasColors": { @@ -1997,11 +2013,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 15, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -2020,12 +2039,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "avg(irate(container_network_transmit_bytes_total{cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", + "expr": "avg(irate(container_network_transmit_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{namespace}}", @@ -2041,7 +2060,7 @@ data: "title": "Average Container Bandwidth by Namespace: Transmitted", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -2078,7 +2097,7 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Average Container Bandwidth by Namespace", "titleSize": "h6" }, { @@ -2095,11 +2114,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 16, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -2118,12 +2140,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(irate(container_network_receive_packets_total{cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", + "expr": "sum(irate(container_network_receive_packets_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{namespace}}", @@ -2139,7 +2161,7 @@ data: "title": "Rate of Received Packets", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -2154,7 +2176,7 @@ data: }, "yaxes": [ { - "format": "Bps", + "format": "pps", "label": null, "logBase": 1, "max": null, @@ -2170,19 +2192,7 @@ data: "show": false } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Network", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ + }, { "aliasColors": { @@ -2193,11 +2203,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 17, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -2216,12 +2229,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(irate(container_network_transmit_packets_total{cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", + "expr": "sum(irate(container_network_transmit_packets_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{namespace}}", @@ -2237,7 +2250,7 @@ data: "title": "Rate of Transmitted Packets", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -2252,7 +2265,7 @@ data: }, "yaxes": [ { - "format": "Bps", + "format": "pps", "label": null, "logBase": 1, "max": null, @@ -2274,7 +2287,7 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Rate of Packets", "titleSize": "h6" }, { @@ -2291,11 +2304,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 18, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -2314,12 +2330,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(irate(container_network_receive_packets_dropped_total{cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", + "expr": "sum(irate(container_network_receive_packets_dropped_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{namespace}}", @@ -2335,7 +2351,7 @@ data: "title": "Rate of Received Packets Dropped", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -2350,7 +2366,7 @@ data: }, "yaxes": [ { - "format": "Bps", + "format": "pps", "label": null, "logBase": 1, "max": null, @@ -2366,19 +2382,7 @@ data: "show": false } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Network", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ + }, { "aliasColors": { @@ -2389,11 +2393,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 19, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -2412,12 +2419,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(irate(container_network_transmit_packets_dropped_total{cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", + "expr": "sum(irate(container_network_transmit_packets_dropped_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=~\".+\"}[$__rate_interval])) by (namespace)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{namespace}}", @@ -2433,7 +2440,198 @@ data: "title": "Rate of Transmitted Packets Dropped", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Rate of Packets Dropped", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "decimals": -1, + "fill": 10, + "id": 20, + "interval": "1m", + "legend": { + "alignAsTable": true, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 0, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "ceil(sum by(namespace) (rate(container_fs_reads_total{job=\"kubernetes-cadvisor\", container!=\"\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", cluster=\"$cluster\", namespace!=\"\"}[$__rate_interval]) + rate(container_fs_writes_total{job=\"kubernetes-cadvisor\", container!=\"\", cluster=\"$cluster\", namespace!=\"\"}[$__rate_interval])))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{namespace}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "IOPS(Reads+Writes)", + "tooltip": { + "shared": false, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 21, + "interval": "1m", + "legend": { + "alignAsTable": true, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 0, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum by(namespace) (rate(container_fs_reads_bytes_total{job=\"kubernetes-cadvisor\", container!=\"\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", cluster=\"$cluster\", namespace!=\"\"}[$__rate_interval]) + rate(container_fs_writes_bytes_total{job=\"kubernetes-cadvisor\", container!=\"\", cluster=\"$cluster\", namespace!=\"\"}[$__rate_interval]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{namespace}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "ThroughPut(Read+Write)", + "tooltip": { + "shared": false, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -2470,7 +2668,315 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Storage IO", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 22, + "interval": "1m", + "legend": { + "alignAsTable": true, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "sort": { + "col": 4, + "desc": true + }, + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": false, + "styles": [ + { + "alias": "Time", + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "pattern": "Time", + "type": "hidden" + }, + { + "alias": "IOPS(Reads)", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": -1, + "link": false, + "linkTargetBlank": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #A", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "IOPS(Writes)", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": -1, + "link": false, + "linkTargetBlank": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #B", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "IOPS(Reads + Writes)", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": -1, + "link": false, + "linkTargetBlank": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #C", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "Throughput(Read)", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTargetBlank": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #D", + "thresholds": [ + + ], + "type": "number", + "unit": "Bps" + }, + { + "alias": "Throughput(Write)", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTargetBlank": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #E", + "thresholds": [ + + ], + "type": "number", + "unit": "Bps" + }, + { + "alias": "Throughput(Read + Write)", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTargetBlank": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #F", + "thresholds": [ + + ], + "type": "number", + "unit": "Bps" + }, + { + "alias": "Namespace", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": true, + "linkTargetBlank": false, + "linkTooltip": "Drill down to pods", + "linkUrl": "/d/85a562078cdf77779eaa1add43ccec1e/k8s-resources-namespace?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$__cell", + "pattern": "namespace", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "pattern": "/.*/", + "thresholds": [ + + ], + "type": "string", + "unit": "short" + } + ], + "targets": [ + { + "expr": "sum by(namespace) (rate(container_fs_reads_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace!=\"\"}[$__rate_interval]))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "A", + "step": 10 + }, + { + "expr": "sum by(namespace) (rate(container_fs_writes_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace!=\"\"}[$__rate_interval]))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "B", + "step": 10 + }, + { + "expr": "sum by(namespace) (rate(container_fs_reads_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace!=\"\"}[$__rate_interval]) + rate(container_fs_writes_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace!=\"\"}[$__rate_interval]))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "C", + "step": 10 + }, + { + "expr": "sum by(namespace) (rate(container_fs_reads_bytes_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace!=\"\"}[$__rate_interval]))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "D", + "step": 10 + }, + { + "expr": "sum by(namespace) (rate(container_fs_writes_bytes_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace!=\"\"}[$__rate_interval]))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "E", + "step": 10 + }, + { + "expr": "sum by(namespace) (rate(container_fs_reads_bytes_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace!=\"\"}[$__rate_interval]) + rate(container_fs_writes_bytes_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace!=\"\"}[$__rate_interval]))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "F", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Current Storage IO", + "tooltip": { + "shared": false, + "sort": 2, + "value_type": "individual" + }, + "transform": "table", + "type": "table", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Storage IO - Distribution", "titleSize": "h6" } ], @@ -2487,7 +2993,7 @@ data: "value": "default" }, "hide": 0, - "label": null, + "label": "Data Source", "name": "datasource", "options": [ @@ -2512,7 +3018,7 @@ data: "options": [ ], - "query": "label_values(node_cpu_seconds_total, cluster)", + "query": "label_values(up{job=\"kubernetes-cadvisor\"}, cluster)", "refresh": 2, "regex": "", "sort": 1, @@ -2591,11 +3097,14 @@ data: "fill": 1, "format": "percentunit", "id": 1, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -2619,7 +3128,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"})", + "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", namespace=\"$namespace\"}) / sum(kube_pod_container_resource_requests{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"cpu\"})", "format": "time_series", "instant": true, "intervalFactor": 2, @@ -2632,7 +3141,7 @@ data: "title": "CPU Utilisation (from requests)", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "singlestat", @@ -2675,11 +3184,14 @@ data: "fill": 1, "format": "percentunit", "id": 2, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -2703,7 +3215,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"})", + "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", namespace=\"$namespace\"}) / sum(kube_pod_container_resource_limits{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"cpu\"})", "format": "time_series", "instant": true, "intervalFactor": 2, @@ -2716,7 +3228,7 @@ data: "title": "CPU Utilisation (from limits)", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "singlestat", @@ -2759,11 +3271,14 @@ data: "fill": 1, "format": "percentunit", "id": 3, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -2787,7 +3302,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\", image!=\"\"}) / sum(kube_pod_container_resource_requests_memory_bytes{namespace=\"$namespace\"})", + "expr": "sum(container_memory_working_set_bytes{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\", image!=\"\"}) / sum(kube_pod_container_resource_requests{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"memory\"})", "format": "time_series", "instant": true, "intervalFactor": 2, @@ -2797,10 +3312,10 @@ data: "thresholds": "70,80", "timeFrom": null, "timeShift": null, - "title": "Memory Utilization (from requests)", + "title": "Memory Utilisation (from requests)", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "singlestat", @@ -2843,11 +3358,14 @@ data: "fill": 1, "format": "percentunit", "id": 4, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -2871,7 +3389,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\", image!=\"\"}) / sum(kube_pod_container_resource_limits_memory_bytes{namespace=\"$namespace\"})", + "expr": "sum(container_memory_working_set_bytes{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\", image!=\"\"}) / sum(kube_pod_container_resource_limits{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"memory\"})", "format": "time_series", "instant": true, "intervalFactor": 2, @@ -2884,7 +3402,7 @@ data: "title": "Memory Utilisation (from limits)", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "singlestat", @@ -2938,11 +3456,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 5, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -2963,8 +3484,9 @@ data: "color": "#F2495C", "dashes": true, "fill": 0, + "hiddenSeries": true, "hideTooltip": true, - "legend": false, + "legend": true, "linewidth": 2, "stack": false }, @@ -2973,8 +3495,9 @@ data: "color": "#FF9830", "dashes": true, "fill": 0, + "hiddenSeries": true, "hideTooltip": true, - "legend": false, + "legend": true, "linewidth": 2, "stack": false } @@ -2985,7 +3508,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", + "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -3017,7 +3540,7 @@ data: "title": "CPU Usage", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -3071,11 +3594,14 @@ data: "datasource": "$datasource", "fill": 1, "id": 6, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -3210,7 +3736,7 @@ data: "link": true, "linkTargetBlank": false, "linkTooltip": "Drill down", - "linkUrl": "./d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell", + "linkUrl": "/d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell", "pattern": "pod", "thresholds": [ @@ -3236,7 +3762,7 @@ data: ], "targets": [ { - "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", + "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -3245,7 +3771,7 @@ data: "step": 10 }, { - "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", + "expr": "sum(cluster:namespace:pod_cpu:active:kube_pod_container_resource_requests{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -3254,7 +3780,7 @@ data: "step": 10 }, { - "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", + "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod) / sum(cluster:namespace:pod_cpu:active:kube_pod_container_resource_requests{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -3263,7 +3789,7 @@ data: "step": 10 }, { - "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", + "expr": "sum(cluster:namespace:pod_cpu:active:kube_pod_container_resource_limits{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -3272,7 +3798,7 @@ data: "step": 10 }, { - "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", + "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod) / sum(cluster:namespace:pod_cpu:active:kube_pod_container_resource_limits{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -3289,7 +3815,7 @@ data: "title": "CPU Quota", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "transform": "table", @@ -3344,11 +3870,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 7, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -3369,8 +3898,9 @@ data: "color": "#F2495C", "dashes": true, "fill": 0, + "hiddenSeries": true, "hideTooltip": true, - "legend": false, + "legend": true, "linewidth": 2, "stack": false }, @@ -3379,8 +3909,9 @@ data: "color": "#FF9830", "dashes": true, "fill": 0, + "hiddenSeries": true, "hideTooltip": true, - "legend": false, + "legend": true, "linewidth": 2, "stack": false } @@ -3391,7 +3922,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\", image!=\"\"}) by (pod)", + "expr": "sum(container_memory_working_set_bytes{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\", image!=\"\"}) by (pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -3423,7 +3954,7 @@ data: "title": "Memory Usage (w/o cache)", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -3477,11 +4008,14 @@ data: "datasource": "$datasource", "fill": 1, "id": 8, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -3673,7 +4207,7 @@ data: "link": true, "linkTargetBlank": false, "linkTooltip": "Drill down", - "linkUrl": "./d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell", + "linkUrl": "/d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell", "pattern": "pod", "thresholds": [ @@ -3699,7 +4233,7 @@ data: ], "targets": [ { - "expr": "sum(container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\", image!=\"\"}) by (pod)", + "expr": "sum(container_memory_working_set_bytes{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\", image!=\"\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -3708,7 +4242,7 @@ data: "step": 10 }, { - "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", + "expr": "sum(cluster:namespace:pod_memory:active:kube_pod_container_resource_requests{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -3717,7 +4251,7 @@ data: "step": 10 }, { - "expr": "sum(container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\", image!=\"\"}) by (pod) / sum(kube_pod_container_resource_requests_memory_bytes{namespace=\"$namespace\"}) by (pod)", + "expr": "sum(container_memory_working_set_bytes{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\", image!=\"\"}) by (pod) / sum(cluster:namespace:pod_memory:active:kube_pod_container_resource_requests{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -3726,7 +4260,7 @@ data: "step": 10 }, { - "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", + "expr": "sum(cluster:namespace:pod_memory:active:kube_pod_container_resource_limits{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -3735,7 +4269,7 @@ data: "step": 10 }, { - "expr": "sum(container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\", image!=\"\"}) by (pod) / sum(kube_pod_container_resource_limits_memory_bytes{namespace=\"$namespace\"}) by (pod)", + "expr": "sum(container_memory_working_set_bytes{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\", image!=\"\"}) by (pod) / sum(cluster:namespace:pod_memory:active:kube_pod_container_resource_limits{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -3744,7 +4278,7 @@ data: "step": 10 }, { - "expr": "sum(container_memory_rss{cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\"}) by (pod)", + "expr": "sum(container_memory_rss{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -3753,7 +4287,7 @@ data: "step": 10 }, { - "expr": "sum(container_memory_cache{cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\"}) by (pod)", + "expr": "sum(container_memory_cache{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -3762,7 +4296,7 @@ data: "step": 10 }, { - "expr": "sum(container_memory_swap{cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\"}) by (pod)", + "expr": "sum(container_memory_swap{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -3779,7 +4313,7 @@ data: "title": "Memory Quota", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "transform": "table", @@ -3836,10 +4370,12 @@ data: "id": 9, "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -3993,7 +4529,7 @@ data: "link": true, "linkTargetBlank": false, "linkTooltip": "Drill down to pods", - "linkUrl": "./d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell", + "linkUrl": "/d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell", "pattern": "pod", "thresholds": [ @@ -4019,7 +4555,7 @@ data: ], "targets": [ { - "expr": "sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])) by (pod)", + "expr": "sum(irate(container_network_receive_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -4028,7 +4564,7 @@ data: "step": 10 }, { - "expr": "sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])) by (pod)", + "expr": "sum(irate(container_network_transmit_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -4037,7 +4573,7 @@ data: "step": 10 }, { - "expr": "sum(irate(container_network_receive_packets_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])) by (pod)", + "expr": "sum(irate(container_network_receive_packets_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -4046,7 +4582,7 @@ data: "step": 10 }, { - "expr": "sum(irate(container_network_transmit_packets_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])) by (pod)", + "expr": "sum(irate(container_network_transmit_packets_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -4055,7 +4591,7 @@ data: "step": 10 }, { - "expr": "sum(irate(container_network_receive_packets_dropped_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])) by (pod)", + "expr": "sum(irate(container_network_receive_packets_dropped_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -4064,7 +4600,7 @@ data: "step": 10 }, { - "expr": "sum(irate(container_network_transmit_packets_dropped_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])) by (pod)", + "expr": "sum(irate(container_network_transmit_packets_dropped_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -4081,7 +4617,7 @@ data: "title": "Current Network Usage", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "transform": "table", @@ -4119,7 +4655,7 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Current Network Usage", "titleSize": "h6" }, { @@ -4136,11 +4672,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 10, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -4159,12 +4698,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])) by (pod)", + "expr": "sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])) by (pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -4180,7 +4719,7 @@ data: "title": "Receive Bandwidth", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -4211,19 +4750,7 @@ data: "show": false } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Network", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ + }, { "aliasColors": { @@ -4234,11 +4761,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 11, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -4257,12 +4787,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])) by (pod)", + "expr": "sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])) by (pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -4278,7 +4808,7 @@ data: "title": "Transmit Bandwidth", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -4315,7 +4845,7 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Bandwidth", "titleSize": "h6" }, { @@ -4332,11 +4862,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 12, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -4355,12 +4888,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(irate(container_network_receive_packets_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])) by (pod)", + "expr": "sum(irate(container_network_receive_packets_total{cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])) by (pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -4376,7 +4909,7 @@ data: "title": "Rate of Received Packets", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -4391,7 +4924,7 @@ data: }, "yaxes": [ { - "format": "Bps", + "format": "pps", "label": null, "logBase": 1, "max": null, @@ -4407,19 +4940,7 @@ data: "show": false } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Network", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ + }, { "aliasColors": { @@ -4430,11 +4951,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 13, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -4453,12 +4977,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(irate(container_network_transmit_packets_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])) by (pod)", + "expr": "sum(irate(container_network_transmit_packets_total{cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])) by (pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -4474,7 +4998,7 @@ data: "title": "Rate of Transmitted Packets", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -4489,7 +5013,7 @@ data: }, "yaxes": [ { - "format": "Bps", + "format": "pps", "label": null, "logBase": 1, "max": null, @@ -4511,7 +5035,7 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Rate of Packets", "titleSize": "h6" }, { @@ -4528,11 +5052,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 14, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -4551,12 +5078,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(irate(container_network_receive_packets_dropped_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])) by (pod)", + "expr": "sum(irate(container_network_receive_packets_dropped_total{cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])) by (pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -4572,7 +5099,7 @@ data: "title": "Rate of Received Packets Dropped", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -4587,7 +5114,7 @@ data: }, "yaxes": [ { - "format": "Bps", + "format": "pps", "label": null, "logBase": 1, "max": null, @@ -4603,19 +5130,7 @@ data: "show": false } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Network", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ + }, { "aliasColors": { @@ -4626,11 +5141,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 15, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -4649,12 +5167,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(irate(container_network_transmit_packets_dropped_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])) by (pod)", + "expr": "sum(irate(container_network_transmit_packets_dropped_total{cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])) by (pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -4670,7 +5188,198 @@ data: "title": "Rate of Transmitted Packets Dropped", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Rate of Packets Dropped", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "decimals": -1, + "fill": 10, + "id": 16, + "interval": "1m", + "legend": { + "alignAsTable": true, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 0, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "ceil(sum by(pod) (rate(container_fs_reads_total{container!=\"\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval]) + rate(container_fs_writes_total{container!=\"\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{pod}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "IOPS(Reads+Writes)", + "tooltip": { + "shared": false, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 17, + "interval": "1m", + "legend": { + "alignAsTable": true, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 0, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum by(pod) (rate(container_fs_reads_bytes_total{container!=\"\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval]) + rate(container_fs_writes_bytes_total{container!=\"\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{pod}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "ThroughPut(Read+Write)", + "tooltip": { + "shared": false, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -4707,7 +5416,315 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Storage IO", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 18, + "interval": "1m", + "legend": { + "alignAsTable": true, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "sort": { + "col": 4, + "desc": true + }, + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": false, + "styles": [ + { + "alias": "Time", + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "pattern": "Time", + "type": "hidden" + }, + { + "alias": "IOPS(Reads)", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": -1, + "link": false, + "linkTargetBlank": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #A", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "IOPS(Writes)", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": -1, + "link": false, + "linkTargetBlank": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #B", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "IOPS(Reads + Writes)", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": -1, + "link": false, + "linkTargetBlank": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #C", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "Throughput(Read)", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTargetBlank": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #D", + "thresholds": [ + + ], + "type": "number", + "unit": "Bps" + }, + { + "alias": "Throughput(Write)", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTargetBlank": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #E", + "thresholds": [ + + ], + "type": "number", + "unit": "Bps" + }, + { + "alias": "Throughput(Read + Write)", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTargetBlank": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #F", + "thresholds": [ + + ], + "type": "number", + "unit": "Bps" + }, + { + "alias": "Pod", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": true, + "linkTargetBlank": false, + "linkTooltip": "Drill down to pods", + "linkUrl": "/d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell", + "pattern": "pod", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "pattern": "/.*/", + "thresholds": [ + + ], + "type": "string", + "unit": "short" + } + ], + "targets": [ + { + "expr": "sum by(pod) (rate(container_fs_reads_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval]))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "A", + "step": 10 + }, + { + "expr": "sum by(pod) (rate(container_fs_writes_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval]))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "B", + "step": 10 + }, + { + "expr": "sum by(pod) (rate(container_fs_reads_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval]) + rate(container_fs_writes_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval]))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "C", + "step": 10 + }, + { + "expr": "sum by(pod) (rate(container_fs_reads_bytes_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval]))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "D", + "step": 10 + }, + { + "expr": "sum by(pod) (rate(container_fs_writes_bytes_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval]))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "E", + "step": 10 + }, + { + "expr": "sum by(pod) (rate(container_fs_reads_bytes_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval]) + rate(container_fs_writes_bytes_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval]))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "F", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Current Storage IO", + "tooltip": { + "shared": false, + "sort": 2, + "value_type": "individual" + }, + "transform": "table", + "type": "table", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Storage IO - Distribution", "titleSize": "h6" } ], @@ -4724,7 +5741,7 @@ data: "value": "default" }, "hide": 0, - "label": null, + "label": "Data Source", "name": "datasource", "options": [ @@ -4749,8 +5766,8 @@ data: "options": [ ], - "query": "label_values(kube_pod_info, cluster)", - "refresh": 1, + "query": "label_values(up{job=\"kube-state-metrics\"}, cluster)", + "refresh": 2, "regex": "", "sort": 1, "tagValuesQuery": "", @@ -4776,8 +5793,8 @@ data: "options": [ ], - "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, namespace)", - "refresh": 1, + "query": "label_values(kube_namespace_status_phase{job=\"kube-state-metrics\", cluster=\"$cluster\"}, namespace)", + "refresh": 2, "regex": "", "sort": 1, "tagValuesQuery": "", @@ -4854,11 +5871,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 1, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -4874,7 +5894,17 @@ data: "points": false, "renderer": "flot", "seriesOverrides": [ - + { + "alias": "max capacity", + "color": "#F2495C", + "dashes": true, + "fill": 0, + "hiddenSeries": true, + "hideTooltip": true, + "legend": true, + "linewidth": 2, + "stack": false + } ], "spaceLength": 10, "span": 12, @@ -4882,7 +5912,15 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", node=~\"$node\"}) by (pod)", + "expr": "sum(kube_node_status_capacity{cluster=\"$cluster\", node=~\"$node\", resource=\"cpu\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "max capacity", + "legendLink": null, + "step": 10 + }, + { + "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", node=~\"$node\"}) by (pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -4898,7 +5936,7 @@ data: "title": "CPU Usage", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -4952,11 +5990,14 @@ data: "datasource": "$datasource", "fill": 1, "id": 2, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -5117,7 +6158,7 @@ data: ], "targets": [ { - "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", node=~\"$node\"}) by (pod)", + "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", node=~\"$node\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -5126,7 +6167,7 @@ data: "step": 10 }, { - "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", node=~\"$node\"}) by (pod)", + "expr": "sum(cluster:namespace:pod_cpu:active:kube_pod_container_resource_requests{cluster=\"$cluster\", node=~\"$node\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -5135,7 +6176,7 @@ data: "step": 10 }, { - "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", node=~\"$node\"}) by (pod) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", node=~\"$node\"}) by (pod)", + "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", node=~\"$node\"}) by (pod) / sum(cluster:namespace:pod_cpu:active:kube_pod_container_resource_requests{cluster=\"$cluster\", node=~\"$node\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -5144,7 +6185,7 @@ data: "step": 10 }, { - "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", node=~\"$node\"}) by (pod)", + "expr": "sum(cluster:namespace:pod_cpu:active:kube_pod_container_resource_limits{cluster=\"$cluster\", node=~\"$node\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -5153,7 +6194,7 @@ data: "step": 10 }, { - "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", node=~\"$node\"}) by (pod) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", node=~\"$node\"}) by (pod)", + "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", node=~\"$node\"}) by (pod) / sum(cluster:namespace:pod_cpu:active:kube_pod_container_resource_limits{cluster=\"$cluster\", node=~\"$node\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -5170,7 +6211,7 @@ data: "title": "CPU Quota", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "transform": "table", @@ -5225,11 +6266,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 3, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -5245,13 +6289,31 @@ data: "points": false, "renderer": "flot", "seriesOverrides": [ - + { + "alias": "max capacity", + "color": "#F2495C", + "dashes": true, + "fill": 0, + "hiddenSeries": true, + "hideTooltip": true, + "legend": true, + "linewidth": 2, + "stack": false + } ], "spaceLength": 10, "span": 12, "stack": true, "steppedLine": false, "targets": [ + { + "expr": "sum(kube_node_status_capacity{cluster=\"$cluster\", node=~\"$node\", resource=\"memory\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "max capacity", + "legendLink": null, + "step": 10 + }, { "expr": "sum(node_namespace_pod_container:container_memory_working_set_bytes{cluster=\"$cluster\", node=~\"$node\", container!=\"\"}) by (pod)", "format": "time_series", @@ -5269,7 +6331,7 @@ data: "title": "Memory Usage (w/o cache)", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -5323,11 +6385,14 @@ data: "datasource": "$datasource", "fill": 1, "id": 4, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -5554,7 +6619,7 @@ data: "step": 10 }, { - "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", node=~\"$node\"}) by (pod)", + "expr": "sum(cluster:namespace:pod_memory:active:kube_pod_container_resource_requests{cluster=\"$cluster\", node=~\"$node\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -5563,7 +6628,7 @@ data: "step": 10 }, { - "expr": "sum(node_namespace_pod_container:container_memory_working_set_bytes{cluster=\"$cluster\", node=~\"$node\",container!=\"\"}) by (pod) / sum(kube_pod_container_resource_requests_memory_bytes{node=~\"$node\"}) by (pod)", + "expr": "sum(node_namespace_pod_container:container_memory_working_set_bytes{cluster=\"$cluster\", node=~\"$node\",container!=\"\"}) by (pod) / sum(cluster:namespace:pod_memory:active:kube_pod_container_resource_requests{cluster=\"$cluster\", node=~\"$node\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -5572,7 +6637,7 @@ data: "step": 10 }, { - "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", node=~\"$node\"}) by (pod)", + "expr": "sum(cluster:namespace:pod_memory:active:kube_pod_container_resource_limits{cluster=\"$cluster\", node=~\"$node\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -5581,7 +6646,7 @@ data: "step": 10 }, { - "expr": "sum(node_namespace_pod_container:container_memory_working_set_bytes{cluster=\"$cluster\", node=~\"$node\",container!=\"\"}) by (pod) / sum(kube_pod_container_resource_limits_memory_bytes{node=~\"$node\"}) by (pod)", + "expr": "sum(node_namespace_pod_container:container_memory_working_set_bytes{cluster=\"$cluster\", node=~\"$node\",container!=\"\"}) by (pod) / sum(cluster:namespace:pod_memory:active:kube_pod_container_resource_limits{cluster=\"$cluster\", node=~\"$node\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -5625,7 +6690,7 @@ data: "title": "Memory Quota", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "transform": "table", @@ -5680,7 +6745,7 @@ data: "value": "default" }, "hide": 0, - "label": null, + "label": "Data Source", "name": "datasource", "options": [ @@ -5705,8 +6770,8 @@ data: "options": [ ], - "query": "label_values(kube_pod_info, cluster)", - "refresh": 1, + "query": "label_values(up{job=\"kube-state-metrics\"}, cluster)", + "refresh": 2, "regex": "", "sort": 1, "tagValuesQuery": "", @@ -5732,8 +6797,8 @@ data: "options": [ ], - "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, node)", - "refresh": 1, + "query": "label_values(kube_node_info{cluster=\"$cluster\"}, node)", + "refresh": 2, "regex": "", "sort": 1, "tagValuesQuery": "", diff --git a/addons/grafana/dashboards-k8s-resources-2.yaml b/addons/grafana/dashboards-k8s-resources-2.yaml index 78a34d73..18fe314a 100644 --- a/addons/grafana/dashboards-k8s-resources-2.yaml +++ b/addons/grafana/dashboards-k8s-resources-2.yaml @@ -30,11 +30,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 1, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -75,7 +78,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", cluster=\"$cluster\"}) by (container)", + "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{namespace=\"$namespace\", pod=\"$pod\", cluster=\"$cluster\"}) by (container)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{container}}", @@ -83,7 +86,7 @@ data: "step": 10 }, { - "expr": "sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"})\n", + "expr": "sum(\n kube_pod_container_resource_requests{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", resource=\"cpu\"}\n)\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "requests", @@ -91,7 +94,7 @@ data: "step": 10 }, { - "expr": "sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"})\n", + "expr": "sum(\n kube_pod_container_resource_limits{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", resource=\"cpu\"}\n)\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "limits", @@ -107,7 +110,7 @@ data: "title": "CPU Usage", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -161,11 +164,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 2, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": true, "max": true, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -189,7 +195,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(increase(container_cpu_cfs_throttled_periods_total{namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", container!=\"\", cluster=\"$cluster\"}[5m])) by (container) /sum(increase(container_cpu_cfs_periods_total{namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", container!=\"\", cluster=\"$cluster\"}[5m])) by (container)", + "expr": "sum(increase(container_cpu_cfs_throttled_periods_total{job=\"kubernetes-cadvisor\", namespace=\"$namespace\", pod=\"$pod\", container!=\"\", cluster=\"$cluster\"}[$__rate_interval])) by (container) /sum(increase(container_cpu_cfs_periods_total{job=\"kubernetes-cadvisor\", namespace=\"$namespace\", pod=\"$pod\", container!=\"\", cluster=\"$cluster\"}[$__rate_interval])) by (container)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{container}}", @@ -212,7 +218,7 @@ data: "title": "CPU Throttling", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -266,11 +272,14 @@ data: "datasource": "$datasource", "fill": 1, "id": 3, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -431,7 +440,7 @@ data: ], "targets": [ { - "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\"}) by (container)", + "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, @@ -440,7 +449,7 @@ data: "step": 10 }, { - "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", + "expr": "sum(cluster:namespace:pod_cpu:active:kube_pod_container_resource_requests{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, @@ -449,7 +458,7 @@ data: "step": 10 }, { - "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", + "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container) / sum(cluster:namespace:pod_cpu:active:kube_pod_container_resource_requests{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, @@ -458,7 +467,7 @@ data: "step": 10 }, { - "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", + "expr": "sum(cluster:namespace:pod_cpu:active:kube_pod_container_resource_limits{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, @@ -467,7 +476,7 @@ data: "step": 10 }, { - "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", + "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container) / sum(cluster:namespace:pod_cpu:active:kube_pod_container_resource_limits{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, @@ -484,7 +493,7 @@ data: "title": "CPU Quota", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "transform": "table", @@ -539,11 +548,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 4, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -565,7 +577,7 @@ data: "dashes": true, "fill": 0, "hideTooltip": true, - "legend": false, + "legend": true, "linewidth": 2, "stack": false }, @@ -575,7 +587,7 @@ data: "dashes": true, "fill": 0, "hideTooltip": true, - "legend": false, + "legend": true, "linewidth": 2, "stack": false } @@ -586,7 +598,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", container!=\"\", image!=\"\"}) by (container)", + "expr": "sum(container_memory_working_set_bytes{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"\", image!=\"\"}) by (container)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{container}}", @@ -594,7 +606,7 @@ data: "step": 10 }, { - "expr": "sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"})\n", + "expr": "sum(\n kube_pod_container_resource_requests{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", resource=\"memory\"}\n)\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "requests", @@ -602,7 +614,7 @@ data: "step": 10 }, { - "expr": "sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"})\n", + "expr": "sum(\n kube_pod_container_resource_limits{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", resource=\"memory\"}\n)\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "limits", @@ -615,10 +627,10 @@ data: ], "timeFrom": null, "timeShift": null, - "title": "Memory Usage", + "title": "Memory Usage (WSS)", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -672,11 +684,14 @@ data: "datasource": "$datasource", "fill": 1, "id": 5, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -706,7 +721,7 @@ data: "type": "hidden" }, { - "alias": "Memory Usage", + "alias": "Memory Usage (WSS)", "colorMode": null, "colors": [ @@ -894,7 +909,7 @@ data: ], "targets": [ { - "expr": "sum(container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", container!=\"\", image!=\"\"}) by (container)", + "expr": "sum(container_memory_working_set_bytes{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"\", image!=\"\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, @@ -903,7 +918,7 @@ data: "step": 10 }, { - "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", + "expr": "sum(cluster:namespace:pod_memory:active:kube_pod_container_resource_requests{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, @@ -912,7 +927,7 @@ data: "step": 10 }, { - "expr": "sum(container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", image!=\"\"}) by (container) / sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", + "expr": "sum(container_memory_working_set_bytes{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", image!=\"\"}) by (container) / sum(cluster:namespace:pod_memory:active:kube_pod_container_resource_requests{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, @@ -921,7 +936,7 @@ data: "step": 10 }, { - "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"\"}) by (container)", + "expr": "sum(cluster:namespace:pod_memory:active:kube_pod_container_resource_limits{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, @@ -930,7 +945,7 @@ data: "step": 10 }, { - "expr": "sum(container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"\", image!=\"\"}) by (container) / sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", + "expr": "sum(container_memory_working_set_bytes{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"\", image!=\"\"}) by (container) / sum(cluster:namespace:pod_memory:active:kube_pod_container_resource_limits{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, @@ -939,7 +954,7 @@ data: "step": 10 }, { - "expr": "sum(container_memory_rss{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container != \"\", container != \"POD\"}) by (container)", + "expr": "sum(container_memory_rss{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container != \"\", container != \"POD\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, @@ -948,7 +963,7 @@ data: "step": 10 }, { - "expr": "sum(container_memory_cache{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container != \"\", container != \"POD\"}) by (container)", + "expr": "sum(container_memory_cache{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container != \"\", container != \"POD\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, @@ -957,7 +972,7 @@ data: "step": 10 }, { - "expr": "sum(container_memory_swap{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container != \"\", container != \"POD\"}) by (container)", + "expr": "sum(container_memory_swap{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container != \"\", container != \"POD\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, @@ -974,7 +989,7 @@ data: "title": "Memory Quota", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "transform": "table", @@ -1031,10 +1046,12 @@ data: "id": 6, "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -1053,12 +1070,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(irate(container_network_receive_bytes_total{namespace=~\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)", + "expr": "sum(irate(container_network_receive_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -1074,7 +1091,7 @@ data: "title": "Receive Bandwidth", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -1105,19 +1122,7 @@ data: "show": false } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Network", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ + }, { "aliasColors": { @@ -1130,10 +1135,12 @@ data: "id": 7, "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -1152,12 +1159,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(irate(container_network_transmit_bytes_total{namespace=~\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)", + "expr": "sum(irate(container_network_transmit_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -1173,7 +1180,7 @@ data: "title": "Transmit Bandwidth", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -1210,7 +1217,7 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Bandwidth", "titleSize": "h6" }, { @@ -1229,10 +1236,12 @@ data: "id": 8, "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -1251,12 +1260,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(irate(container_network_receive_packets_total{namespace=~\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)", + "expr": "sum(irate(container_network_receive_packets_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -1272,7 +1281,7 @@ data: "title": "Rate of Received Packets", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -1287,7 +1296,7 @@ data: }, "yaxes": [ { - "format": "Bps", + "format": "pps", "label": null, "logBase": 1, "max": null, @@ -1303,19 +1312,7 @@ data: "show": false } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Network", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ + }, { "aliasColors": { @@ -1328,10 +1325,12 @@ data: "id": 9, "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -1350,12 +1349,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(irate(container_network_transmit_packets_total{namespace=~\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)", + "expr": "sum(irate(container_network_transmit_packets_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -1371,7 +1370,7 @@ data: "title": "Rate of Transmitted Packets", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -1386,7 +1385,7 @@ data: }, "yaxes": [ { - "format": "Bps", + "format": "pps", "label": null, "logBase": 1, "max": null, @@ -1408,7 +1407,7 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Rate of Packets", "titleSize": "h6" }, { @@ -1427,10 +1426,12 @@ data: "id": 10, "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -1449,12 +1450,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(irate(container_network_receive_packets_dropped_total{namespace=~\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)", + "expr": "sum(irate(container_network_receive_packets_dropped_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -1470,7 +1471,7 @@ data: "title": "Rate of Received Packets Dropped", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -1485,7 +1486,7 @@ data: }, "yaxes": [ { - "format": "Bps", + "format": "pps", "label": null, "logBase": 1, "max": null, @@ -1501,19 +1502,7 @@ data: "show": false } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Network", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ + }, { "aliasColors": { @@ -1526,10 +1515,12 @@ data: "id": 11, "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -1548,12 +1539,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "sum(irate(container_network_transmit_packets_dropped_total{namespace=~\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)", + "expr": "sum(irate(container_network_transmit_packets_dropped_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -1569,7 +1560,214 @@ data: "title": "Rate of Transmitted Packets Dropped", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "pps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Rate of Packets Dropped", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "decimals": -1, + "fill": 10, + "id": 12, + "interval": "1m", + "legend": { + "alignAsTable": true, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 0, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "ceil(sum by(pod) (rate(container_fs_reads_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\", pod=~\"$pod\"}[$__rate_interval])))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "Reads", + "legendLink": null, + "step": 10 + }, + { + "expr": "ceil(sum by(pod) (rate(container_fs_writes_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\",namespace=\"$namespace\", pod=~\"$pod\"}[$__rate_interval])))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "Writes", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "IOPS", + "tooltip": { + "shared": false, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 13, + "interval": "1m", + "legend": { + "alignAsTable": true, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 0, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum by(pod) (rate(container_fs_reads_bytes_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\", pod=~\"$pod\"}[$__rate_interval]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "Reads", + "legendLink": null, + "step": 10 + }, + { + "expr": "sum by(pod) (rate(container_fs_writes_bytes_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\", pod=~\"$pod\"}[$__rate_interval]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "Writes", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "ThroughPut", + "tooltip": { + "shared": false, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -1606,7 +1804,506 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Storage IO - Distribution(Pod - Read & Writes)", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "decimals": -1, + "fill": 10, + "id": 14, + "interval": "1m", + "legend": { + "alignAsTable": true, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 0, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "ceil(sum by(container) (rate(container_fs_reads_total{job=\"kubernetes-cadvisor\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}[$__rate_interval]) + rate(container_fs_writes_total{job=\"kubernetes-cadvisor\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}[$__rate_interval])))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{container}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "IOPS(Reads+Writes)", + "tooltip": { + "shared": false, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 15, + "interval": "1m", + "legend": { + "alignAsTable": true, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 0, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum by(container) (rate(container_fs_reads_bytes_total{job=\"kubernetes-cadvisor\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}[$__rate_interval]) + rate(container_fs_writes_bytes_total{job=\"kubernetes-cadvisor\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}[$__rate_interval]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{container}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "ThroughPut(Read+Write)", + "tooltip": { + "shared": false, + "sort": 2, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Storage IO - Distribution(Containers)", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 16, + "interval": "1m", + "legend": { + "alignAsTable": true, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": true, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "sort": { + "col": 4, + "desc": true + }, + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": false, + "styles": [ + { + "alias": "Time", + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "pattern": "Time", + "type": "hidden" + }, + { + "alias": "IOPS(Reads)", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": -1, + "link": false, + "linkTargetBlank": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #A", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "IOPS(Writes)", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": -1, + "link": false, + "linkTargetBlank": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #B", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "IOPS(Reads + Writes)", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": -1, + "link": false, + "linkTargetBlank": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #C", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "Throughput(Read)", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTargetBlank": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #D", + "thresholds": [ + + ], + "type": "number", + "unit": "Bps" + }, + { + "alias": "Throughput(Write)", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTargetBlank": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #E", + "thresholds": [ + + ], + "type": "number", + "unit": "Bps" + }, + { + "alias": "Throughput(Read + Write)", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTargetBlank": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #F", + "thresholds": [ + + ], + "type": "number", + "unit": "Bps" + }, + { + "alias": "Container", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTargetBlank": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "container", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "pattern": "/.*/", + "thresholds": [ + + ], + "type": "string", + "unit": "short" + } + ], + "targets": [ + { + "expr": "sum by(container) (rate(container_fs_reads_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}[$__rate_interval]))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "A", + "step": 10 + }, + { + "expr": "sum by(container) (rate(container_fs_writes_total{job=\"kubernetes-cadvisor\",device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}[$__rate_interval]))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "B", + "step": 10 + }, + { + "expr": "sum by(container) (rate(container_fs_reads_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}[$__rate_interval]) + rate(container_fs_writes_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}[$__rate_interval]))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "C", + "step": 10 + }, + { + "expr": "sum by(container) (rate(container_fs_reads_bytes_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}[$__rate_interval]))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "D", + "step": 10 + }, + { + "expr": "sum by(container) (rate(container_fs_writes_bytes_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}[$__rate_interval]))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "E", + "step": 10 + }, + { + "expr": "sum by(container) (rate(container_fs_reads_bytes_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}[$__rate_interval]) + rate(container_fs_writes_bytes_total{job=\"kubernetes-cadvisor\", device=~\"mmcblk.p.+|nvme.+|rbd.+|sd.+|vd.+|xvd.+|dm-.+|dasd.+\", container!=\"\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}[$__rate_interval]))", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "F", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Current Storage IO", + "tooltip": { + "shared": false, + "sort": 2, + "value_type": "individual" + }, + "transform": "table", + "type": "table", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Storage IO - Distribution", "titleSize": "h6" } ], @@ -1623,7 +2320,7 @@ data: "value": "default" }, "hide": 0, - "label": null, + "label": "Data Source", "name": "datasource", "options": [ @@ -1648,8 +2345,8 @@ data: "options": [ ], - "query": "label_values(kube_pod_info, cluster)", - "refresh": 1, + "query": "label_values(up{job=\"kube-state-metrics\"}, cluster)", + "refresh": 2, "regex": "", "sort": 1, "tagValuesQuery": "", @@ -1675,8 +2372,8 @@ data: "options": [ ], - "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, namespace)", - "refresh": 1, + "query": "label_values(kube_namespace_status_phase{job=\"kube-state-metrics\", cluster=\"$cluster\"}, namespace)", + "refresh": 2, "regex": "", "sort": 1, "tagValuesQuery": "", @@ -1702,7 +2399,7 @@ data: "options": [ ], - "query": "label_values(kube_pod_info{cluster=\"$cluster\", namespace=\"$namespace\"}, pod)", + "query": "label_values(kube_pod_info{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\"}, pod)", "refresh": 2, "regex": "", "sort": 1, @@ -1780,11 +2477,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 1, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -1808,7 +2508,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -1824,7 +2524,7 @@ data: "title": "CPU Usage", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -1878,11 +2578,14 @@ data: "datasource": "$datasource", "fill": 1, "id": 2, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -2017,7 +2720,7 @@ data: "link": true, "linkTargetBlank": false, "linkTooltip": "Drill down", - "linkUrl": "./d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell", + "linkUrl": "/d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell", "pattern": "pod", "thresholds": [ @@ -2043,7 +2746,7 @@ data: ], "targets": [ { - "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -2052,7 +2755,7 @@ data: "step": 10 }, { - "expr": "sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "expr": "sum(\n kube_pod_container_resource_requests{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"cpu\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -2061,7 +2764,7 @@ data: "step": 10 }, { - "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_requests{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"cpu\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -2070,7 +2773,7 @@ data: "step": 10 }, { - "expr": "sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "expr": "sum(\n kube_pod_container_resource_limits{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"cpu\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -2079,7 +2782,7 @@ data: "step": 10 }, { - "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_limits{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"cpu\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -2096,7 +2799,7 @@ data: "title": "CPU Quota", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "transform": "table", @@ -2151,11 +2854,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 3, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -2195,7 +2901,7 @@ data: "title": "Memory Usage", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -2249,11 +2955,14 @@ data: "datasource": "$datasource", "fill": 1, "id": 4, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -2388,7 +3097,7 @@ data: "link": true, "linkTargetBlank": false, "linkTooltip": "Drill down", - "linkUrl": "./d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell", + "linkUrl": "/d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell", "pattern": "pod", "thresholds": [ @@ -2423,7 +3132,7 @@ data: "step": 10 }, { - "expr": "sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "expr": "sum(\n kube_pod_container_resource_requests{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"memory\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -2432,7 +3141,7 @@ data: "step": 10 }, { - "expr": "sum(\n container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\", image!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "expr": "sum(\n container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\", image!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_requests{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"memory\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -2441,7 +3150,7 @@ data: "step": 10 }, { - "expr": "sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "expr": "sum(\n kube_pod_container_resource_limits{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"memory\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -2450,7 +3159,7 @@ data: "step": 10 }, { - "expr": "sum(\n container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\", image!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "expr": "sum(\n container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\", image!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_limits{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"memory\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -2467,7 +3176,7 @@ data: "title": "Memory Quota", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "transform": "table", @@ -2524,10 +3233,12 @@ data: "id": 5, "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -2681,7 +3392,7 @@ data: "link": true, "linkTargetBlank": false, "linkTooltip": "Drill down", - "linkUrl": "./d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell", + "linkUrl": "/d/6581e46e4e5c7ba40a07646395ef7b23/k8s-resources-pod?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-pod=$__cell", "pattern": "pod", "thresholds": [ @@ -2707,7 +3418,7 @@ data: ], "targets": [ { - "expr": "(sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "expr": "(sum(irate(container_network_receive_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -2716,7 +3427,7 @@ data: "step": 10 }, { - "expr": "(sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "expr": "(sum(irate(container_network_transmit_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -2725,7 +3436,7 @@ data: "step": 10 }, { - "expr": "(sum(irate(container_network_receive_packets_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "expr": "(sum(irate(container_network_receive_packets_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -2734,7 +3445,7 @@ data: "step": 10 }, { - "expr": "(sum(irate(container_network_transmit_packets_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "expr": "(sum(irate(container_network_transmit_packets_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -2743,7 +3454,7 @@ data: "step": 10 }, { - "expr": "(sum(irate(container_network_receive_packets_dropped_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "expr": "(sum(irate(container_network_receive_packets_dropped_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -2752,7 +3463,7 @@ data: "step": 10 }, { - "expr": "(sum(irate(container_network_transmit_packets_dropped_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "expr": "(sum(irate(container_network_transmit_packets_dropped_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -2769,7 +3480,7 @@ data: "title": "Current Network Usage", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "transform": "table", @@ -2807,7 +3518,7 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Current Network Usage", "titleSize": "h6" }, { @@ -2824,11 +3535,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 6, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -2847,12 +3561,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "(sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "expr": "(sum(irate(container_network_receive_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -2868,7 +3582,7 @@ data: "title": "Receive Bandwidth", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -2899,19 +3613,7 @@ data: "show": false } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Network", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ + }, { "aliasColors": { @@ -2922,11 +3624,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 7, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -2945,12 +3650,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "(sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "expr": "(sum(irate(container_network_transmit_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -2966,7 +3671,7 @@ data: "title": "Transmit Bandwidth", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -3003,7 +3708,7 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Bandwidth", "titleSize": "h6" }, { @@ -3020,11 +3725,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 8, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -3043,12 +3751,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "(avg(irate(container_network_receive_bytes_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "expr": "(avg(irate(container_network_receive_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -3064,7 +3772,7 @@ data: "title": "Average Container Bandwidth by Pod: Received", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -3095,19 +3803,7 @@ data: "show": false } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Network", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ + }, { "aliasColors": { @@ -3118,11 +3814,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 9, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -3141,12 +3840,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "(avg(irate(container_network_transmit_bytes_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "expr": "(avg(irate(container_network_transmit_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -3162,7 +3861,7 @@ data: "title": "Average Container Bandwidth by Pod: Transmitted", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -3199,7 +3898,7 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Average Container Bandwidth by Pod", "titleSize": "h6" }, { @@ -3216,11 +3915,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 10, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -3239,12 +3941,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "(sum(irate(container_network_receive_packets_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "expr": "(sum(irate(container_network_receive_packets_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -3260,7 +3962,7 @@ data: "title": "Rate of Received Packets", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -3275,7 +3977,7 @@ data: }, "yaxes": [ { - "format": "Bps", + "format": "pps", "label": null, "logBase": 1, "max": null, @@ -3291,19 +3993,7 @@ data: "show": false } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Network", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ + }, { "aliasColors": { @@ -3314,11 +4004,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 11, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -3337,12 +4030,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "(sum(irate(container_network_transmit_packets_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "expr": "(sum(irate(container_network_transmit_packets_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -3358,7 +4051,7 @@ data: "title": "Rate of Transmitted Packets", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -3373,7 +4066,7 @@ data: }, "yaxes": [ { - "format": "Bps", + "format": "pps", "label": null, "logBase": 1, "max": null, @@ -3395,7 +4088,7 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Rate of Packets", "titleSize": "h6" }, { @@ -3412,11 +4105,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 12, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -3435,12 +4131,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "(sum(irate(container_network_receive_packets_dropped_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "expr": "(sum(irate(container_network_receive_packets_dropped_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -3456,7 +4152,7 @@ data: "title": "Rate of Received Packets Dropped", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -3471,7 +4167,7 @@ data: }, "yaxes": [ { - "format": "Bps", + "format": "pps", "label": null, "logBase": 1, "max": null, @@ -3487,19 +4183,7 @@ data: "show": false } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Network", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ + }, { "aliasColors": { @@ -3510,11 +4194,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 13, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -3533,12 +4220,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "(sum(irate(container_network_transmit_packets_dropped_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", + "expr": "(sum(irate(container_network_transmit_packets_dropped_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\"$workload\", workload_type=\"$type\"}) by (pod))\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -3554,7 +4241,7 @@ data: "title": "Rate of Transmitted Packets Dropped", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -3569,7 +4256,7 @@ data: }, "yaxes": [ { - "format": "Bps", + "format": "pps", "label": null, "logBase": 1, "max": null, @@ -3591,7 +4278,7 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Rate of Packets Dropped", "titleSize": "h6" } ], @@ -3608,7 +4295,7 @@ data: "value": "default" }, "hide": 0, - "label": null, + "label": "Data Source", "name": "datasource", "options": [ @@ -3633,8 +4320,8 @@ data: "options": [ ], - "query": "label_values(kube_pod_info, cluster)", - "refresh": 1, + "query": "label_values(up{job=\"kube-state-metrics\"}, cluster)", + "refresh": 2, "regex": "", "sort": 1, "tagValuesQuery": "", @@ -3660,35 +4347,8 @@ data: "options": [ ], - "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, namespace)", - "refresh": 1, - "regex": "", - "sort": 1, - "tagValuesQuery": "", - "tags": [ - - ], - "tagsQuery": "", - "type": "query", - "useTags": false - }, - { - "allValue": null, - "current": { - "text": "", - "value": "" - }, - "datasource": "$datasource", - "hide": 0, - "includeAll": false, - "label": null, - "multi": false, - "name": "workload", - "options": [ - - ], - "query": "label_values(namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\"}, workload)", - "refresh": 1, + "query": "label_values(kube_namespace_status_phase{job=\"kube-state-metrics\", cluster=\"$cluster\"}, namespace)", + "refresh": 2, "regex": "", "sort": 1, "tagValuesQuery": "", @@ -3714,8 +4374,35 @@ data: "options": [ ], - "query": "label_values(namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\"}, workload_type)", - "refresh": 1, + "query": "label_values(namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\"}, workload_type)", + "refresh": 2, + "regex": "", + "sort": 1, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "current": { + "text": "", + "value": "" + }, + "datasource": "$datasource", + "hide": 0, + "includeAll": false, + "label": null, + "multi": false, + "name": "workload", + "options": [ + + ], + "query": "label_values(namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}, workload)", + "refresh": 2, "regex": "", "sort": 1, "tagValuesQuery": "", @@ -3792,11 +4479,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 1, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -3817,8 +4507,9 @@ data: "color": "#F2495C", "dashes": true, "fill": 0, + "hiddenSeries": true, "hideTooltip": true, - "legend": false, + "legend": true, "linewidth": 2, "stack": false }, @@ -3827,8 +4518,9 @@ data: "color": "#FF9830", "dashes": true, "fill": 0, + "hiddenSeries": true, "hideTooltip": true, - "legend": false, + "legend": true, "linewidth": 2, "stack": false } @@ -3839,7 +4531,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{workload}} - {{workload_type}}", @@ -3871,7 +4563,7 @@ data: "title": "CPU Usage", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -3925,11 +4617,14 @@ data: "datasource": "$datasource", "fill": 1, "id": 2, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -4083,7 +4778,7 @@ data: "link": true, "linkTargetBlank": false, "linkTooltip": "Drill down", - "linkUrl": "./d/a164a7f0339f99e89cea5cb47e9be617/k8s-resources-workload?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-workload=$__cell&var-type=$__cell_2", + "linkUrl": "/d/a164a7f0339f99e89cea5cb47e9be617/k8s-resources-workload?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-workload=$__cell&var-type=$__cell_2", "pattern": "workload", "thresholds": [ @@ -4137,7 +4832,7 @@ data: "step": 10 }, { - "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4146,7 +4841,7 @@ data: "step": 10 }, { - "expr": "sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n kube_pod_container_resource_requests{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"cpu\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4155,7 +4850,7 @@ data: "step": 10 }, { - "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_requests{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"cpu\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4164,7 +4859,7 @@ data: "step": 10 }, { - "expr": "sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n kube_pod_container_resource_limits{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"cpu\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4173,7 +4868,7 @@ data: "step": 10 }, { - "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_limits{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"cpu\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4190,7 +4885,7 @@ data: "title": "CPU Quota", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "transform": "table", @@ -4245,11 +4940,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 3, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -4270,8 +4968,9 @@ data: "color": "#F2495C", "dashes": true, "fill": 0, + "hiddenSeries": true, "hideTooltip": true, - "legend": false, + "legend": true, "linewidth": 2, "stack": false }, @@ -4280,8 +4979,9 @@ data: "color": "#FF9830", "dashes": true, "fill": 0, + "hiddenSeries": true, "hideTooltip": true, - "legend": false, + "legend": true, "linewidth": 2, "stack": false } @@ -4292,7 +4992,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(\n container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\", image!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n container_memory_working_set_bytes{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\", image!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{workload}} - {{workload_type}}", @@ -4324,7 +5024,7 @@ data: "title": "Memory Usage", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -4378,11 +5078,14 @@ data: "datasource": "$datasource", "fill": 1, "id": 4, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -4536,7 +5239,7 @@ data: "link": true, "linkTargetBlank": false, "linkTooltip": "Drill down", - "linkUrl": "./d/a164a7f0339f99e89cea5cb47e9be617/k8s-resources-workload?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-workload=$__cell&var-type=$__cell_2", + "linkUrl": "/d/a164a7f0339f99e89cea5cb47e9be617/k8s-resources-workload?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-workload=$__cell&var-type=$__cell_2", "pattern": "workload", "thresholds": [ @@ -4590,7 +5293,7 @@ data: "step": 10 }, { - "expr": "sum(\n container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\", image!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n container_memory_working_set_bytes{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\", image!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4599,7 +5302,7 @@ data: "step": 10 }, { - "expr": "sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n kube_pod_container_resource_requests{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"memory\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4608,7 +5311,7 @@ data: "step": 10 }, { - "expr": "sum(\n container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\", image!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n container_memory_working_set_bytes{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\", image!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_requests{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"memory\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4617,7 +5320,7 @@ data: "step": 10 }, { - "expr": "sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n kube_pod_container_resource_limits{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"memory\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4626,7 +5329,7 @@ data: "step": 10 }, { - "expr": "sum(\n container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\", image!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n container_memory_working_set_bytes{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\", image!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_limits{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", resource=\"memory\"}\n* on(namespace,pod)\n group_left(workload, workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}\n) by (workload, workload_type)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4643,7 +5346,7 @@ data: "title": "Memory Quota", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "transform": "table", @@ -4700,10 +5403,12 @@ data: "id": 5, "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -4857,7 +5562,7 @@ data: "link": true, "linkTargetBlank": false, "linkTooltip": "Drill down to pods", - "linkUrl": "./d/a164a7f0339f99e89cea5cb47e9be617/k8s-resources-workload?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-workload=$__cell&var-type=$type", + "linkUrl": "/d/a164a7f0339f99e89cea5cb47e9be617/k8s-resources-workload?var-datasource=$datasource&var-cluster=$cluster&var-namespace=$namespace&var-workload=$__cell&var-type=$type", "pattern": "workload", "thresholds": [ @@ -4902,7 +5607,7 @@ data: ], "targets": [ { - "expr": "(sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload_type=\"$type\"}) by (workload))\n", + "expr": "(sum(irate(container_network_receive_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}) by (workload))\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4911,7 +5616,7 @@ data: "step": 10 }, { - "expr": "(sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload_type=\"$type\"}) by (workload))\n", + "expr": "(sum(irate(container_network_transmit_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}) by (workload))\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4920,7 +5625,7 @@ data: "step": 10 }, { - "expr": "(sum(irate(container_network_receive_packets_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload_type=\"$type\"}) by (workload))\n", + "expr": "(sum(irate(container_network_receive_packets_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}) by (workload))\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4929,7 +5634,7 @@ data: "step": 10 }, { - "expr": "(sum(irate(container_network_transmit_packets_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload_type=\"$type\"}) by (workload))\n", + "expr": "(sum(irate(container_network_transmit_packets_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}) by (workload))\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4938,7 +5643,7 @@ data: "step": 10 }, { - "expr": "(sum(irate(container_network_receive_packets_dropped_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload_type=\"$type\"}) by (workload))\n", + "expr": "(sum(irate(container_network_receive_packets_dropped_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}) by (workload))\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4947,7 +5652,7 @@ data: "step": 10 }, { - "expr": "(sum(irate(container_network_transmit_packets_dropped_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload_type=\"$type\"}) by (workload))\n", + "expr": "(sum(irate(container_network_transmit_packets_dropped_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload_type=\"$type\"}) by (workload))\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4964,7 +5669,7 @@ data: "title": "Current Network Usage", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "transform": "table", @@ -5002,7 +5707,7 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Current Network Usage", "titleSize": "h6" }, { @@ -5019,11 +5724,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 6, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -5042,12 +5750,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "(sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "expr": "(sum(irate(container_network_receive_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{workload}}", @@ -5063,7 +5771,7 @@ data: "title": "Receive Bandwidth", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -5094,19 +5802,7 @@ data: "show": false } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Network", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ + }, { "aliasColors": { @@ -5117,11 +5813,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 7, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -5140,12 +5839,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "(sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "expr": "(sum(irate(container_network_transmit_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{workload}}", @@ -5161,7 +5860,7 @@ data: "title": "Transmit Bandwidth", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -5198,7 +5897,7 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Bandwidth", "titleSize": "h6" }, { @@ -5215,11 +5914,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 8, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -5238,12 +5940,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "(avg(irate(container_network_receive_bytes_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "expr": "(avg(irate(container_network_receive_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{workload}}", @@ -5259,7 +5961,7 @@ data: "title": "Average Container Bandwidth by Workload: Received", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -5290,19 +5992,7 @@ data: "show": false } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Network", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ + }, { "aliasColors": { @@ -5313,11 +6003,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 9, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -5336,12 +6029,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "(avg(irate(container_network_transmit_bytes_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "expr": "(avg(irate(container_network_transmit_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{workload}}", @@ -5357,7 +6050,7 @@ data: "title": "Average Container Bandwidth by Workload: Transmitted", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -5394,7 +6087,7 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Average Container Bandwidth by Workload", "titleSize": "h6" }, { @@ -5411,11 +6104,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 10, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -5434,12 +6130,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "(sum(irate(container_network_receive_packets_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "expr": "(sum(irate(container_network_receive_packets_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{workload}}", @@ -5455,7 +6151,7 @@ data: "title": "Rate of Received Packets", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -5470,7 +6166,7 @@ data: }, "yaxes": [ { - "format": "Bps", + "format": "pps", "label": null, "logBase": 1, "max": null, @@ -5486,19 +6182,7 @@ data: "show": false } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Network", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ + }, { "aliasColors": { @@ -5509,11 +6193,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 11, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -5532,12 +6219,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "(sum(irate(container_network_transmit_packets_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "expr": "(sum(irate(container_network_transmit_packets_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{workload}}", @@ -5553,7 +6240,7 @@ data: "title": "Rate of Transmitted Packets", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -5568,7 +6255,7 @@ data: }, "yaxes": [ { - "format": "Bps", + "format": "pps", "label": null, "logBase": 1, "max": null, @@ -5590,7 +6277,7 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Rate of Packets", "titleSize": "h6" }, { @@ -5607,11 +6294,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 12, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -5630,12 +6320,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "(sum(irate(container_network_receive_packets_dropped_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "expr": "(sum(irate(container_network_receive_packets_dropped_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{workload}}", @@ -5651,7 +6341,7 @@ data: "title": "Rate of Received Packets Dropped", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -5666,7 +6356,7 @@ data: }, "yaxes": [ { - "format": "Bps", + "format": "pps", "label": null, "logBase": 1, "max": null, @@ -5682,19 +6372,7 @@ data: "show": false } ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Network", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ + }, { "aliasColors": { @@ -5705,11 +6383,14 @@ data: "datasource": "$datasource", "fill": 10, "id": 13, + "interval": "1m", "legend": { + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, + "rightSide": true, "show": true, "total": false, "values": false @@ -5728,12 +6409,12 @@ data: ], "spaceLength": 10, - "span": 12, + "span": 6, "stack": true, "steppedLine": false, "targets": [ { - "expr": "(sum(irate(container_network_transmit_packets_dropped_total{cluster=\"$cluster\", namespace=~\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", + "expr": "(sum(irate(container_network_transmit_packets_dropped_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\"}[$__rate_interval])\n* on (namespace,pod)\ngroup_left(workload,workload_type) namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\".+\", workload_type=\"$type\"}) by (workload))\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{workload}}", @@ -5749,7 +6430,7 @@ data: "title": "Rate of Transmitted Packets Dropped", "tooltip": { "shared": false, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -5764,7 +6445,7 @@ data: }, "yaxes": [ { - "format": "Bps", + "format": "pps", "label": null, "logBase": 1, "max": null, @@ -5786,7 +6467,7 @@ data: "repeatIteration": null, "repeatRowId": null, "showTitle": true, - "title": "Network", + "title": "Rate of Packets Dropped", "titleSize": "h6" } ], @@ -5803,7 +6484,7 @@ data: "value": "default" }, "hide": 0, - "label": null, + "label": "Data Source", "name": "datasource", "options": [ @@ -5813,38 +6494,6 @@ data: "regex": "", "type": "datasource" }, - { - "allValue": null, - "auto": false, - "auto_count": 30, - "auto_min": "10s", - "current": { - "text": "deployment", - "value": "deployment" - }, - "datasource": "$datasource", - "definition": "label_values(namespace_workload_pod:kube_pod_owner:relabel{namespace=~\"$namespace\", workload=~\".+\"}, workload_type)", - "hide": 0, - "includeAll": false, - "label": null, - "multi": false, - "name": "type", - "options": [ - - ], - "query": "label_values(namespace_workload_pod:kube_pod_owner:relabel{namespace=~\"$namespace\", workload=~\".+\"}, workload_type)", - "refresh": 1, - "regex": "", - "skipUrlSync": false, - "sort": 0, - "tagValuesQuery": "", - "tags": [ - - ], - "tagsQuery": "", - "type": "query", - "useTags": false - }, { "allValue": null, "current": { @@ -5860,8 +6509,8 @@ data: "options": [ ], - "query": "label_values(kube_pod_info, cluster)", - "refresh": 1, + "query": "label_values(up{job=\"kube-state-metrics\"}, cluster)", + "refresh": 2, "regex": "", "sort": 1, "tagValuesQuery": "", @@ -5887,13 +6536,45 @@ data: "options": [ ], - "query": "label_values(kube_pod_info{cluster=\"$cluster\"}, namespace)", - "refresh": 1, + "query": "label_values(kube_pod_info{job=\"kube-state-metrics\", cluster=\"$cluster\"}, namespace)", + "refresh": 2, "regex": "", "sort": 1, "tagValuesQuery": "", "tags": [ + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "auto": false, + "auto_count": 30, + "auto_min": "10s", + "current": { + "text": "deployment", + "value": "deployment" + }, + "datasource": "$datasource", + "definition": "label_values(namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\".+\"}, workload_type)", + "hide": 0, + "includeAll": false, + "label": null, + "multi": false, + "name": "type", + "options": [ + + ], + "query": "label_values(namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=\"$namespace\", workload=~\".+\"}, workload_type)", + "refresh": 2, + "regex": "", + "skipUrlSync": false, + "sort": 0, + "tagValuesQuery": "", + "tags": [ + ], "tagsQuery": "", "type": "query", diff --git a/addons/grafana/dashboards-k8s.yaml b/addons/grafana/dashboards-k8s.yaml index e5714fde..dc6ead4e 100644 --- a/addons/grafana/dashboards-k8s.yaml +++ b/addons/grafana/dashboards-k8s.yaml @@ -69,7 +69,11 @@ data: }, "id": 3, - "interval": null, + "interval": "1m", + "legend": { + "alignAsTable": true, + "rightSide": true + }, "links": [ ], @@ -147,13 +151,14 @@ data: }, "id": 4, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -265,7 +270,11 @@ data: }, "id": 5, - "interval": null, + "interval": "1m", + "legend": { + "alignAsTable": true, + "rightSide": true + }, "links": [ ], @@ -342,13 +351,14 @@ data: }, "id": 6, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -451,13 +461,14 @@ data: }, "id": 7, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -545,13 +556,14 @@ data: }, "id": 8, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -661,7 +673,11 @@ data: }, "id": 9, - "interval": null, + "interval": "1m", + "legend": { + "alignAsTable": true, + "rightSide": true + }, "links": [ ], @@ -738,13 +754,14 @@ data: }, "id": 10, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -847,13 +864,14 @@ data: }, "id": 11, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -941,13 +959,14 @@ data: }, "id": 12, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -1047,13 +1066,14 @@ data: }, "id": 13, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": false, "sideWidth": null, "total": false, @@ -1079,7 +1099,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(rate(workqueue_adds_total{job=\"apiserver\", instance=~\"$instance\", cluster=\"$cluster\"}[5m])) by (instance, name)", + "expr": "sum(rate(workqueue_adds_total{job=\"apiserver\", instance=~\"$instance\", cluster=\"$cluster\"}[$__rate_interval])) by (instance, name)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{instance}} {{name}}", @@ -1140,13 +1160,14 @@ data: }, "id": 14, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": false, "sideWidth": null, "total": false, @@ -1172,7 +1193,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(rate(workqueue_depth{job=\"apiserver\", instance=~\"$instance\", cluster=\"$cluster\"}[5m])) by (instance, name)", + "expr": "sum(rate(workqueue_depth{job=\"apiserver\", instance=~\"$instance\", cluster=\"$cluster\"}[$__rate_interval])) by (instance, name)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{instance}} {{name}}", @@ -1233,6 +1254,7 @@ data: }, "id": 15, + "interval": "1m", "legend": { "alignAsTable": true, "avg": false, @@ -1265,7 +1287,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "histogram_quantile(0.99, sum(rate(workqueue_queue_duration_seconds_bucket{job=\"apiserver\", instance=~\"$instance\", cluster=\"$cluster\"}[5m])) by (instance, name, le))", + "expr": "histogram_quantile(0.99, sum(rate(workqueue_queue_duration_seconds_bucket{job=\"apiserver\", instance=~\"$instance\", cluster=\"$cluster\"}[$__rate_interval])) by (instance, name, le))", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{instance}} {{name}}", @@ -1339,13 +1361,14 @@ data: }, "id": 16, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -1432,13 +1455,14 @@ data: }, "id": 17, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -1464,7 +1488,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "rate(process_cpu_seconds_total{job=\"apiserver\",instance=~\"$instance\", cluster=\"$cluster\"}[5m])", + "expr": "rate(process_cpu_seconds_total{job=\"apiserver\",instance=~\"$instance\", cluster=\"$cluster\"}[$__rate_interval])", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{instance}}", @@ -1525,13 +1549,14 @@ data: }, "id": 18, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -1627,7 +1652,7 @@ data: "value": "default" }, "hide": 0, - "label": null, + "label": "Data Source", "name": "datasource", "options": [ @@ -1651,7 +1676,7 @@ data: "options": [ ], - "query": "label_values(apiserver_request_total, cluster)", + "query": "label_values(up{job=\"apiserver\"}, cluster)", "refresh": 2, "regex": "", "sort": 1, @@ -1677,7 +1702,7 @@ data: "options": [ ], - "query": "label_values(apiserver_request_total{job=\"apiserver\", cluster=\"$cluster\"}, instance)", + "query": "label_values(up{job=\"apiserver\", cluster=\"$cluster\"}, instance)", "refresh": 2, "regex": "", "sort": 1, @@ -1774,7 +1799,11 @@ data: }, "id": 2, - "interval": null, + "interval": "1m", + "legend": { + "alignAsTable": true, + "rightSide": true + }, "links": [ ], @@ -1813,7 +1842,7 @@ data: "tableColumn": "", "targets": [ { - "expr": "sum(up{job=\"kube-controller-manager\"})", + "expr": "sum(up{cluster=\"$cluster\", job=\"kube-controller-manager\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "", @@ -1850,6 +1879,7 @@ data: }, "id": 3, + "interval": "1m", "legend": { "alignAsTable": true, "avg": false, @@ -1882,10 +1912,10 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(rate(workqueue_adds_total{job=\"kube-controller-manager\", instance=~\"$instance\"}[5m])) by (instance, name)", + "expr": "sum(rate(workqueue_adds_total{cluster=\"$cluster\", job=\"kube-controller-manager\", instance=~\"$instance\"}[$__rate_interval])) by (cluster, instance, name)", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{instance}} {{name}}", + "legendFormat": "{{cluster}} {{instance}} {{name}}", "refId": "A" } ], @@ -1956,6 +1986,7 @@ data: }, "id": 4, + "interval": "1m", "legend": { "alignAsTable": true, "avg": false, @@ -1988,10 +2019,10 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(rate(workqueue_depth{job=\"kube-controller-manager\", instance=~\"$instance\"}[5m])) by (instance, name)", + "expr": "sum(rate(workqueue_depth{cluster=\"$cluster\", job=\"kube-controller-manager\", instance=~\"$instance\"}[$__rate_interval])) by (cluster, instance, name)", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{instance}} {{name}}", + "legendFormat": "{{cluster}} {{instance}} {{name}}", "refId": "A" } ], @@ -2062,6 +2093,7 @@ data: }, "id": 5, + "interval": "1m", "legend": { "alignAsTable": true, "avg": false, @@ -2094,10 +2126,10 @@ data: "steppedLine": false, "targets": [ { - "expr": "histogram_quantile(0.99, sum(rate(workqueue_queue_duration_seconds_bucket{job=\"kube-controller-manager\", instance=~\"$instance\"}[5m])) by (instance, name, le))", + "expr": "histogram_quantile(0.99, sum(rate(workqueue_queue_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-controller-manager\", instance=~\"$instance\"}[$__rate_interval])) by (cluster, instance, name, le))", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{instance}} {{name}}", + "legendFormat": "{{cluster}} {{instance}} {{name}}", "refId": "A" } ], @@ -2168,13 +2200,14 @@ data: }, "id": 6, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -2200,28 +2233,28 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(rate(rest_client_requests_total{job=\"kube-controller-manager\", instance=~\"$instance\",code=~\"2..\"}[5m]))", + "expr": "sum(rate(rest_client_requests_total{job=\"kube-controller-manager\", instance=~\"$instance\",code=~\"2..\"}[$__rate_interval]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "2xx", "refId": "A" }, { - "expr": "sum(rate(rest_client_requests_total{job=\"kube-controller-manager\", instance=~\"$instance\",code=~\"3..\"}[5m]))", + "expr": "sum(rate(rest_client_requests_total{job=\"kube-controller-manager\", instance=~\"$instance\",code=~\"3..\"}[$__rate_interval]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "3xx", "refId": "B" }, { - "expr": "sum(rate(rest_client_requests_total{job=\"kube-controller-manager\", instance=~\"$instance\",code=~\"4..\"}[5m]))", + "expr": "sum(rate(rest_client_requests_total{job=\"kube-controller-manager\", instance=~\"$instance\",code=~\"4..\"}[$__rate_interval]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "4xx", "refId": "C" }, { - "expr": "sum(rate(rest_client_requests_total{job=\"kube-controller-manager\", instance=~\"$instance\",code=~\"5..\"}[5m]))", + "expr": "sum(rate(rest_client_requests_total{job=\"kube-controller-manager\", instance=~\"$instance\",code=~\"5..\"}[$__rate_interval]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "5xx", @@ -2282,13 +2315,14 @@ data: }, "id": 7, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -2314,7 +2348,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{job=\"kube-controller-manager\", instance=~\"$instance\", verb=\"POST\"}[5m])) by (verb, url, le))", + "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-controller-manager\", instance=~\"$instance\", verb=\"POST\"}[$__rate_interval])) by (verb, url, le))", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{verb}} {{url}}", @@ -2388,6 +2422,7 @@ data: }, "id": 8, + "interval": "1m", "legend": { "alignAsTable": true, "avg": false, @@ -2420,7 +2455,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{job=\"kube-controller-manager\", instance=~\"$instance\", verb=\"GET\"}[5m])) by (verb, url, le))", + "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-controller-manager\", instance=~\"$instance\", verb=\"GET\"}[$__rate_interval])) by (verb, url, le))", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{verb}} {{url}}", @@ -2494,13 +2529,14 @@ data: }, "id": 9, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -2526,7 +2562,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "process_resident_memory_bytes{job=\"kube-controller-manager\",instance=~\"$instance\"}", + "expr": "process_resident_memory_bytes{cluster=\"$cluster\", job=\"kube-controller-manager\",instance=~\"$instance\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{instance}}", @@ -2587,13 +2623,14 @@ data: }, "id": 10, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -2619,7 +2656,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "rate(process_cpu_seconds_total{job=\"kube-controller-manager\",instance=~\"$instance\"}[5m])", + "expr": "rate(process_cpu_seconds_total{cluster=\"$cluster\", job=\"kube-controller-manager\",instance=~\"$instance\"}[$__rate_interval])", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{instance}}", @@ -2680,13 +2717,14 @@ data: }, "id": 11, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -2712,7 +2750,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "go_goroutines{job=\"kube-controller-manager\",instance=~\"$instance\"}", + "expr": "go_goroutines{cluster=\"$cluster\", job=\"kube-controller-manager\",instance=~\"$instance\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{instance}}", @@ -2782,7 +2820,7 @@ data: "value": "default" }, "hide": 0, - "label": null, + "label": "Data Source", "name": "datasource", "options": [ @@ -2796,6 +2834,32 @@ data: "allValue": null, "current": { + }, + "datasource": "$datasource", + "hide": 2, + "includeAll": false, + "label": "cluster", + "multi": false, + "name": "cluster", + "options": [ + + ], + "query": "label_values(up{job=\"kube-controller-manager\"}, cluster)", + "refresh": 2, + "regex": "", + "sort": 1, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "current": { + }, "datasource": "$datasource", "hide": 0, @@ -2806,7 +2870,7 @@ data: "options": [ ], - "query": "label_values(process_cpu_seconds_total{job=\"kube-controller-manager\"}, instance)", + "query": "label_values(up{cluster=\"$cluster\", job=\"kube-controller-manager\"}, instance)", "refresh": 2, "regex": "", "sort": 1, @@ -2895,13 +2959,14 @@ data: }, "id": 2, + "interval": "1m", "legend": { "alignAsTable": true, "avg": true, "current": true, "max": true, "min": true, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -2927,14 +2992,14 @@ data: "steppedLine": false, "targets": [ { - "expr": "(\n sum without(instance, node) (kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n -\n sum without(instance, node) (kubelet_volume_stats_available_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n)\n", + "expr": "(\n sum without(instance, node) (topk(1, (kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})))\n -\n sum without(instance, node) (topk(1, (kubelet_volume_stats_available_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})))\n)\n", "format": "time_series", "intervalFactor": 1, "legendFormat": "Used Space", "refId": "A" }, { - "expr": "sum without(instance, node) (kubelet_volume_stats_available_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n", + "expr": "sum without(instance, node) (topk(1, (kubelet_volume_stats_available_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})))\n", "format": "time_series", "intervalFactor": 1, "legendFormat": "Free Space", @@ -3003,7 +3068,11 @@ data: }, "id": 3, - "interval": null, + "interval": "1m", + "legend": { + "alignAsTable": true, + "rightSide": true + }, "links": [ ], @@ -3042,7 +3111,7 @@ data: "tableColumn": "", "targets": [ { - "expr": "max without(instance,node) (\n(\n kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n -\n kubelet_volume_stats_available_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n)\n/\nkubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n* 100)\n", + "expr": "max without(instance,node) (\n(\n topk(1, kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n -\n topk(1, kubelet_volume_stats_available_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n)\n/\ntopk(1, kubelet_volume_stats_capacity_bytes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n* 100)\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "", @@ -3092,13 +3161,14 @@ data: }, "id": 4, + "interval": "1m", "legend": { "alignAsTable": true, "avg": true, "current": true, "max": true, "min": true, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -3124,14 +3194,14 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum without(instance, node) (kubelet_volume_stats_inodes_used{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n", + "expr": "sum without(instance, node) (topk(1, (kubelet_volume_stats_inodes_used{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})))\n", "format": "time_series", "intervalFactor": 1, "legendFormat": "Used inodes", "refId": "A" }, { - "expr": "(\n sum without(instance, node) (kubelet_volume_stats_inodes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n -\n sum without(instance, node) (kubelet_volume_stats_inodes_used{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n)\n", + "expr": "(\n sum without(instance, node) (topk(1, (kubelet_volume_stats_inodes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})))\n -\n sum without(instance, node) (topk(1, (kubelet_volume_stats_inodes_used{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})))\n)\n", "format": "time_series", "intervalFactor": 1, "legendFormat": " Free inodes", @@ -3200,7 +3270,11 @@ data: }, "id": 5, - "interval": null, + "interval": "1m", + "legend": { + "alignAsTable": true, + "rightSide": true + }, "links": [ ], @@ -3239,7 +3313,7 @@ data: "tableColumn": "", "targets": [ { - "expr": "max without(instance,node) (\nkubelet_volume_stats_inodes_used{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n/\nkubelet_volume_stats_inodes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"}\n* 100)\n", + "expr": "max without(instance,node) (\ntopk(1, kubelet_volume_stats_inodes_used{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n/\ntopk(1, kubelet_volume_stats_inodes{cluster=\"$cluster\", job=\"kubelet\", namespace=\"$namespace\", persistentvolumeclaim=\"$volume\"})\n* 100)\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "", @@ -3285,7 +3359,7 @@ data: "value": "default" }, "hide": 0, - "label": null, + "label": "Data Source", "name": "datasource", "options": [ @@ -3309,7 +3383,7 @@ data: "options": [ ], - "query": "label_values(kubelet_volume_stats_capacity_bytes, cluster)", + "query": "label_values(kubelet_volume_stats_capacity_bytes{job=\"kubelet\"}, cluster)", "refresh": 2, "regex": "", "sort": 1, @@ -3458,7 +3532,11 @@ data: }, "id": 2, - "interval": null, + "interval": "1m", + "legend": { + "alignAsTable": true, + "rightSide": true + }, "links": [ ], @@ -3497,7 +3575,7 @@ data: "tableColumn": "", "targets": [ { - "expr": "sum(up{job=\"kube-scheduler\"})", + "expr": "sum(up{cluster=\"$cluster\", job=\"kube-scheduler\"})", "format": "time_series", "intervalFactor": 2, "legendFormat": "", @@ -3534,6 +3612,7 @@ data: }, "id": 3, + "interval": "1m", "legend": { "alignAsTable": true, "avg": false, @@ -3566,31 +3645,31 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(rate(scheduler_e2e_scheduling_duration_seconds_count{job=\"kube-scheduler\", instance=~\"$instance\"}[5m])) by (instance)", + "expr": "sum(rate(scheduler_e2e_scheduling_duration_seconds_count{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\"}[$__rate_interval])) by (cluster, instance)", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{instance}} e2e", + "legendFormat": "{{cluster}} {{instance}} e2e", "refId": "A" }, { - "expr": "sum(rate(scheduler_binding_duration_seconds_count{job=\"kube-scheduler\", instance=~\"$instance\"}[5m])) by (instance)", + "expr": "sum(rate(scheduler_binding_duration_seconds_count{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\"}[$__rate_interval])) by (cluster, instance)", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{instance}} binding", + "legendFormat": "{{cluster}} {{instance}} binding", "refId": "B" }, { - "expr": "sum(rate(scheduler_scheduling_algorithm_duration_seconds_count{job=\"kube-scheduler\", instance=~\"$instance\"}[5m])) by (instance)", + "expr": "sum(rate(scheduler_scheduling_algorithm_duration_seconds_count{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\"}[$__rate_interval])) by (cluster, instance)", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{instance}} scheduling algorithm", + "legendFormat": "{{cluster}} {{instance}} scheduling algorithm", "refId": "C" }, { - "expr": "sum(rate(scheduler_volume_scheduling_duration_seconds_count{job=\"kube-scheduler\", instance=~\"$instance\"}[5m])) by (instance)", + "expr": "sum(rate(scheduler_volume_scheduling_duration_seconds_count{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\"}[$__rate_interval])) by (cluster, instance)", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{instance}} volume", + "legendFormat": "{{cluster}} {{instance}} volume", "refId": "D" } ], @@ -3648,6 +3727,7 @@ data: }, "id": 4, + "interval": "1m", "legend": { "alignAsTable": true, "avg": false, @@ -3680,31 +3760,31 @@ data: "steppedLine": false, "targets": [ { - "expr": "histogram_quantile(0.99, sum(rate(scheduler_e2e_scheduling_duration_seconds_bucket{job=\"kube-scheduler\",instance=~\"$instance\"}[5m])) by (instance, le))", + "expr": "histogram_quantile(0.99, sum(rate(scheduler_e2e_scheduling_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-scheduler\",instance=~\"$instance\"}[$__rate_interval])) by (cluster, instance, le))", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{instance}} e2e", + "legendFormat": "{{cluster}} {{instance}} e2e", "refId": "A" }, { - "expr": "histogram_quantile(0.99, sum(rate(scheduler_binding_duration_seconds_bucket{job=\"kube-scheduler\",instance=~\"$instance\"}[5m])) by (instance, le))", + "expr": "histogram_quantile(0.99, sum(rate(scheduler_binding_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-scheduler\",instance=~\"$instance\"}[$__rate_interval])) by (cluster, instance, le))", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{instance}} binding", + "legendFormat": "{{cluster}} {{instance}} binding", "refId": "B" }, { - "expr": "histogram_quantile(0.99, sum(rate(scheduler_scheduling_algorithm_duration_seconds_bucket{job=\"kube-scheduler\",instance=~\"$instance\"}[5m])) by (instance, le))", + "expr": "histogram_quantile(0.99, sum(rate(scheduler_scheduling_algorithm_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-scheduler\",instance=~\"$instance\"}[$__rate_interval])) by (cluster, instance, le))", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{instance}} scheduling algorithm", + "legendFormat": "{{cluster}} {{instance}} scheduling algorithm", "refId": "C" }, { - "expr": "histogram_quantile(0.99, sum(rate(scheduler_volume_scheduling_duration_seconds_bucket{job=\"kube-scheduler\",instance=~\"$instance\"}[5m])) by (instance, le))", + "expr": "histogram_quantile(0.99, sum(rate(scheduler_volume_scheduling_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-scheduler\",instance=~\"$instance\"}[$__rate_interval])) by (cluster, instance, le))", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{instance}} volume", + "legendFormat": "{{cluster}} {{instance}} volume", "refId": "D" } ], @@ -3775,13 +3855,14 @@ data: }, "id": 5, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -3807,28 +3888,28 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(rate(rest_client_requests_total{job=\"kube-scheduler\", instance=~\"$instance\",code=~\"2..\"}[5m]))", + "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\",code=~\"2..\"}[$__rate_interval]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "2xx", "refId": "A" }, { - "expr": "sum(rate(rest_client_requests_total{job=\"kube-scheduler\", instance=~\"$instance\",code=~\"3..\"}[5m]))", + "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\",code=~\"3..\"}[$__rate_interval]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "3xx", "refId": "B" }, { - "expr": "sum(rate(rest_client_requests_total{job=\"kube-scheduler\", instance=~\"$instance\",code=~\"4..\"}[5m]))", + "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\",code=~\"4..\"}[$__rate_interval]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "4xx", "refId": "C" }, { - "expr": "sum(rate(rest_client_requests_total{job=\"kube-scheduler\", instance=~\"$instance\",code=~\"5..\"}[5m]))", + "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\",code=~\"5..\"}[$__rate_interval]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "5xx", @@ -3889,13 +3970,14 @@ data: }, "id": 6, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -3921,7 +4003,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{job=\"kube-scheduler\", instance=~\"$instance\", verb=\"POST\"}[5m])) by (verb, url, le))", + "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\", verb=\"POST\"}[$__rate_interval])) by (verb, url, le))", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{verb}} {{url}}", @@ -3995,6 +4077,7 @@ data: }, "id": 7, + "interval": "1m", "legend": { "alignAsTable": true, "avg": false, @@ -4027,7 +4110,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{job=\"kube-scheduler\", instance=~\"$instance\", verb=\"GET\"}[5m])) by (verb, url, le))", + "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\", verb=\"GET\"}[$__rate_interval])) by (verb, url, le))", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{verb}} {{url}}", @@ -4101,13 +4184,14 @@ data: }, "id": 8, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -4133,7 +4217,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "process_resident_memory_bytes{job=\"kube-scheduler\", instance=~\"$instance\"}", + "expr": "process_resident_memory_bytes{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{instance}}", @@ -4194,13 +4278,14 @@ data: }, "id": 9, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -4226,7 +4311,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "rate(process_cpu_seconds_total{job=\"kube-scheduler\", instance=~\"$instance\"}[5m])", + "expr": "rate(process_cpu_seconds_total{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\"}[$__rate_interval])", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{instance}}", @@ -4287,13 +4372,14 @@ data: }, "id": 10, + "interval": "1m", "legend": { - "alignAsTable": false, + "alignAsTable": true, "avg": false, "current": false, "max": false, "min": false, - "rightSide": false, + "rightSide": true, "show": true, "sideWidth": null, "total": false, @@ -4319,7 +4405,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "go_goroutines{job=\"kube-scheduler\",instance=~\"$instance\"}", + "expr": "go_goroutines{cluster=\"$cluster\", job=\"kube-scheduler\",instance=~\"$instance\"}", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{instance}}", @@ -4389,7 +4475,7 @@ data: "value": "default" }, "hide": 0, - "label": null, + "label": "Data Source", "name": "datasource", "options": [ @@ -4403,6 +4489,32 @@ data: "allValue": null, "current": { + }, + "datasource": "$datasource", + "hide": 2, + "includeAll": false, + "label": "cluster", + "multi": false, + "name": "cluster", + "options": [ + + ], + "query": "label_values(up{job=\"kube-scheduler\"}, cluster)", + "refresh": 2, + "regex": "", + "sort": 1, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "current": { + }, "datasource": "$datasource", "hide": 0, @@ -4413,7 +4525,7 @@ data: "options": [ ], - "query": "label_values(process_cpu_seconds_total{job=\"kube-scheduler\"}, instance)", + "query": "label_values(up{job=\"kube-scheduler\", cluster=\"$cluster\"}, instance)", "refresh": 2, "regex": "", "sort": 1, @@ -4461,912 +4573,6 @@ data: "uid": "2e6b6a3b4bddf1427b3a55aa1311c656", "version": 0 } - statefulset.json: |- - { - "__inputs": [ - - ], - "__requires": [ - - ], - "annotations": { - "list": [ - - ] - }, - "editable": false, - "gnetId": null, - "graphTooltip": 0, - "hideControls": false, - "id": null, - "links": [ - - ], - "refresh": "", - "rows": [ - { - "collapse": false, - "collapsed": false, - "panels": [ - { - "cacheTimeout": null, - "colorBackground": false, - "colorValue": false, - "colors": [ - "#299c46", - "rgba(237, 129, 40, 0.89)", - "#d44a3a" - ], - "datasource": "$datasource", - "format": "none", - "gauge": { - "maxValue": 100, - "minValue": 0, - "show": false, - "thresholdLabels": false, - "thresholdMarkers": true - }, - "gridPos": { - - }, - "id": 2, - "interval": null, - "links": [ - - ], - "mappingType": 1, - "mappingTypes": [ - { - "name": "value to text", - "value": 1 - }, - { - "name": "range to text", - "value": 2 - } - ], - "maxDataPoints": 100, - "nullPointMode": "connected", - "nullText": null, - "postfix": "cores", - "postfixFontSize": "50%", - "prefix": "", - "prefixFontSize": "50%", - "rangeMaps": [ - { - "from": "null", - "text": "N/A", - "to": "null" - } - ], - "span": 4, - "sparkline": { - "fillColor": "rgba(31, 118, 189, 0.18)", - "lineColor": "rgb(31, 120, 193)", - "show": true - }, - "tableColumn": "", - "targets": [ - { - "expr": "sum(rate(container_cpu_usage_seconds_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", container!=\"\", namespace=\"$namespace\", pod=~\"$statefulset.*\"}[3m]))", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "", - "refId": "A" - } - ], - "thresholds": "", - "title": "CPU", - "tooltip": { - "shared": false - }, - "type": "singlestat", - "valueFontSize": "80%", - "valueMaps": [ - { - "op": "=", - "text": "0", - "value": "null" - } - ], - "valueName": "current" - }, - { - "cacheTimeout": null, - "colorBackground": false, - "colorValue": false, - "colors": [ - "#299c46", - "rgba(237, 129, 40, 0.89)", - "#d44a3a" - ], - "datasource": "$datasource", - "format": "none", - "gauge": { - "maxValue": 100, - "minValue": 0, - "show": false, - "thresholdLabels": false, - "thresholdMarkers": true - }, - "gridPos": { - - }, - "id": 3, - "interval": null, - "links": [ - - ], - "mappingType": 1, - "mappingTypes": [ - { - "name": "value to text", - "value": 1 - }, - { - "name": "range to text", - "value": 2 - } - ], - "maxDataPoints": 100, - "nullPointMode": "connected", - "nullText": null, - "postfix": "GB", - "postfixFontSize": "50%", - "prefix": "", - "prefixFontSize": "50%", - "rangeMaps": [ - { - "from": "null", - "text": "N/A", - "to": "null" - } - ], - "span": 4, - "sparkline": { - "fillColor": "rgba(31, 118, 189, 0.18)", - "lineColor": "rgb(31, 120, 193)", - "show": true - }, - "tableColumn": "", - "targets": [ - { - "expr": "sum(container_memory_usage_bytes{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", container!=\"\", namespace=\"$namespace\", pod=~\"$statefulset.*\"}) / 1024^3", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "", - "refId": "A" - } - ], - "thresholds": "", - "title": "Memory", - "tooltip": { - "shared": false - }, - "type": "singlestat", - "valueFontSize": "80%", - "valueMaps": [ - { - "op": "=", - "text": "0", - "value": "null" - } - ], - "valueName": "current" - }, - { - "cacheTimeout": null, - "colorBackground": false, - "colorValue": false, - "colors": [ - "#299c46", - "rgba(237, 129, 40, 0.89)", - "#d44a3a" - ], - "datasource": "$datasource", - "format": "none", - "gauge": { - "maxValue": 100, - "minValue": 0, - "show": false, - "thresholdLabels": false, - "thresholdMarkers": true - }, - "gridPos": { - - }, - "id": 4, - "interval": null, - "links": [ - - ], - "mappingType": 1, - "mappingTypes": [ - { - "name": "value to text", - "value": 1 - }, - { - "name": "range to text", - "value": 2 - } - ], - "maxDataPoints": 100, - "nullPointMode": "connected", - "nullText": null, - "postfix": "Bps", - "postfixFontSize": "50%", - "prefix": "", - "prefixFontSize": "50%", - "rangeMaps": [ - { - "from": "null", - "text": "N/A", - "to": "null" - } - ], - "span": 4, - "sparkline": { - "fillColor": "rgba(31, 118, 189, 0.18)", - "lineColor": "rgb(31, 120, 193)", - "show": true - }, - "tableColumn": "", - "targets": [ - { - "expr": "sum(rate(container_network_transmit_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod=~\"$statefulset.*\"}[3m])) + sum(rate(container_network_receive_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\",pod=~\"$statefulset.*\"}[3m]))", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "", - "refId": "A" - } - ], - "thresholds": "", - "title": "Network", - "tooltip": { - "shared": false - }, - "type": "singlestat", - "valueFontSize": "80%", - "valueMaps": [ - { - "op": "=", - "text": "0", - "value": "null" - } - ], - "valueName": "current" - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Dashboard Row", - "titleSize": "h6", - "type": "row" - }, - { - "collapse": false, - "collapsed": false, - "height": "100px", - "panels": [ - { - "cacheTimeout": null, - "colorBackground": false, - "colorValue": false, - "colors": [ - "#299c46", - "rgba(237, 129, 40, 0.89)", - "#d44a3a" - ], - "datasource": "$datasource", - "format": "none", - "gauge": { - "maxValue": 100, - "minValue": 0, - "show": false, - "thresholdLabels": false, - "thresholdMarkers": true - }, - "gridPos": { - - }, - "id": 5, - "interval": null, - "links": [ - - ], - "mappingType": 1, - "mappingTypes": [ - { - "name": "value to text", - "value": 1 - }, - { - "name": "range to text", - "value": 2 - } - ], - "maxDataPoints": 100, - "nullPointMode": "connected", - "nullText": null, - "postfix": "", - "postfixFontSize": "50%", - "prefix": "", - "prefixFontSize": "50%", - "rangeMaps": [ - { - "from": "null", - "text": "N/A", - "to": "null" - } - ], - "span": 3, - "sparkline": { - "fillColor": "rgba(31, 118, 189, 0.18)", - "full": false, - "lineColor": "rgb(31, 120, 193)", - "show": false - }, - "tableColumn": "", - "targets": [ - { - "expr": "max(kube_statefulset_replicas{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", statefulset=\"$statefulset\"}) without (instance, pod)", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "", - "refId": "A" - } - ], - "thresholds": "", - "title": "Desired Replicas", - "tooltip": { - "shared": false - }, - "type": "singlestat", - "valueFontSize": "80%", - "valueMaps": [ - { - "op": "=", - "text": "0", - "value": "null" - } - ], - "valueName": "current" - }, - { - "cacheTimeout": null, - "colorBackground": false, - "colorValue": false, - "colors": [ - "#299c46", - "rgba(237, 129, 40, 0.89)", - "#d44a3a" - ], - "datasource": "$datasource", - "format": "none", - "gauge": { - "maxValue": 100, - "minValue": 0, - "show": false, - "thresholdLabels": false, - "thresholdMarkers": true - }, - "gridPos": { - - }, - "id": 6, - "interval": null, - "links": [ - - ], - "mappingType": 1, - "mappingTypes": [ - { - "name": "value to text", - "value": 1 - }, - { - "name": "range to text", - "value": 2 - } - ], - "maxDataPoints": 100, - "nullPointMode": "connected", - "nullText": null, - "postfix": "", - "postfixFontSize": "50%", - "prefix": "", - "prefixFontSize": "50%", - "rangeMaps": [ - { - "from": "null", - "text": "N/A", - "to": "null" - } - ], - "span": 3, - "sparkline": { - "fillColor": "rgba(31, 118, 189, 0.18)", - "full": false, - "lineColor": "rgb(31, 120, 193)", - "show": false - }, - "tableColumn": "", - "targets": [ - { - "expr": "min(kube_statefulset_status_replicas_current{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", statefulset=\"$statefulset\"}) without (instance, pod)", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "", - "refId": "A" - } - ], - "thresholds": "", - "title": "Replicas of current version", - "tooltip": { - "shared": false - }, - "type": "singlestat", - "valueFontSize": "80%", - "valueMaps": [ - { - "op": "=", - "text": "0", - "value": "null" - } - ], - "valueName": "current" - }, - { - "cacheTimeout": null, - "colorBackground": false, - "colorValue": false, - "colors": [ - "#299c46", - "rgba(237, 129, 40, 0.89)", - "#d44a3a" - ], - "datasource": "$datasource", - "format": "none", - "gauge": { - "maxValue": 100, - "minValue": 0, - "show": false, - "thresholdLabels": false, - "thresholdMarkers": true - }, - "gridPos": { - - }, - "id": 7, - "interval": null, - "links": [ - - ], - "mappingType": 1, - "mappingTypes": [ - { - "name": "value to text", - "value": 1 - }, - { - "name": "range to text", - "value": 2 - } - ], - "maxDataPoints": 100, - "nullPointMode": "connected", - "nullText": null, - "postfix": "", - "postfixFontSize": "50%", - "prefix": "", - "prefixFontSize": "50%", - "rangeMaps": [ - { - "from": "null", - "text": "N/A", - "to": "null" - } - ], - "span": 3, - "sparkline": { - "fillColor": "rgba(31, 118, 189, 0.18)", - "full": false, - "lineColor": "rgb(31, 120, 193)", - "show": false - }, - "tableColumn": "", - "targets": [ - { - "expr": "max(kube_statefulset_status_observed_generation{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\", statefulset=\"$statefulset\"}) without (instance, pod)", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "", - "refId": "A" - } - ], - "thresholds": "", - "title": "Observed Generation", - "tooltip": { - "shared": false - }, - "type": "singlestat", - "valueFontSize": "80%", - "valueMaps": [ - { - "op": "=", - "text": "0", - "value": "null" - } - ], - "valueName": "current" - }, - { - "cacheTimeout": null, - "colorBackground": false, - "colorValue": false, - "colors": [ - "#299c46", - "rgba(237, 129, 40, 0.89)", - "#d44a3a" - ], - "datasource": "$datasource", - "format": "none", - "gauge": { - "maxValue": 100, - "minValue": 0, - "show": false, - "thresholdLabels": false, - "thresholdMarkers": true - }, - "gridPos": { - - }, - "id": 8, - "interval": null, - "links": [ - - ], - "mappingType": 1, - "mappingTypes": [ - { - "name": "value to text", - "value": 1 - }, - { - "name": "range to text", - "value": 2 - } - ], - "maxDataPoints": 100, - "nullPointMode": "connected", - "nullText": null, - "postfix": "", - "postfixFontSize": "50%", - "prefix": "", - "prefixFontSize": "50%", - "rangeMaps": [ - { - "from": "null", - "text": "N/A", - "to": "null" - } - ], - "span": 3, - "sparkline": { - "fillColor": "rgba(31, 118, 189, 0.18)", - "full": false, - "lineColor": "rgb(31, 120, 193)", - "show": false - }, - "tableColumn": "", - "targets": [ - { - "expr": "max(kube_statefulset_metadata_generation{job=\"kube-state-metrics\", statefulset=\"$statefulset\", cluster=\"$cluster\", namespace=\"$namespace\"}) without (instance, pod)", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "", - "refId": "A" - } - ], - "thresholds": "", - "title": "Metadata Generation", - "tooltip": { - "shared": false - }, - "type": "singlestat", - "valueFontSize": "80%", - "valueMaps": [ - { - "op": "=", - "text": "0", - "value": "null" - } - ], - "valueName": "current" - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Dashboard Row", - "titleSize": "h6", - "type": "row" - }, - { - "collapse": false, - "collapsed": false, - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "fillGradient": 0, - "gridPos": { - - }, - "id": 9, - "legend": { - "alignAsTable": false, - "avg": false, - "current": false, - "max": false, - "min": false, - "rightSide": false, - "show": true, - "sideWidth": null, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - - ], - "spaceLength": 10, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "max(kube_statefulset_replicas{job=\"kube-state-metrics\", statefulset=\"$statefulset\", cluster=\"$cluster\", namespace=\"$namespace\"}) without (instance, pod)", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "replicas specified", - "refId": "A" - }, - { - "expr": "max(kube_statefulset_status_replicas{job=\"kube-state-metrics\", statefulset=\"$statefulset\", cluster=\"$cluster\", namespace=\"$namespace\"}) without (instance, pod)", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "replicas created", - "refId": "B" - }, - { - "expr": "min(kube_statefulset_status_replicas_ready{job=\"kube-state-metrics\", statefulset=\"$statefulset\", cluster=\"$cluster\", namespace=\"$namespace\"}) without (instance, pod)", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "ready", - "refId": "C" - }, - { - "expr": "min(kube_statefulset_status_replicas_current{job=\"kube-state-metrics\", statefulset=\"$statefulset\", cluster=\"$cluster\", namespace=\"$namespace\"}) without (instance, pod)", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "replicas of current version", - "refId": "D" - }, - { - "expr": "min(kube_statefulset_status_replicas_updated{job=\"kube-state-metrics\", statefulset=\"$statefulset\", cluster=\"$cluster\", namespace=\"$namespace\"}) without (instance, pod)", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "updated", - "refId": "E" - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Replicas", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": true - }, - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": true - } - ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": false, - "title": "Dashboard Row", - "titleSize": "h6", - "type": "row" - } - ], - "schemaVersion": 14, - "style": "dark", - "tags": [ - "kubernetes-mixin" - ], - "templating": { - "list": [ - { - "current": { - "text": "default", - "value": "default" - }, - "hide": 0, - "label": null, - "name": "datasource", - "options": [ - - ], - "query": "prometheus", - "refresh": 1, - "regex": "", - "type": "datasource" - }, - { - "allValue": null, - "current": { - - }, - "datasource": "$datasource", - "hide": 2, - "includeAll": false, - "label": "cluster", - "multi": false, - "name": "cluster", - "options": [ - - ], - "query": "label_values(kube_statefulset_metadata_generation, cluster)", - "refresh": 2, - "regex": "", - "sort": 1, - "tagValuesQuery": "", - "tags": [ - - ], - "tagsQuery": "", - "type": "query", - "useTags": false - }, - { - "allValue": null, - "current": { - - }, - "datasource": "$datasource", - "hide": 0, - "includeAll": false, - "label": "Namespace", - "multi": false, - "name": "namespace", - "options": [ - - ], - "query": "label_values(kube_statefulset_metadata_generation{job=\"kube-state-metrics\", cluster=\"$cluster\"}, namespace)", - "refresh": 2, - "regex": "", - "sort": 1, - "tagValuesQuery": "", - "tags": [ - - ], - "tagsQuery": "", - "type": "query", - "useTags": false - }, - { - "allValue": null, - "current": { - - }, - "datasource": "$datasource", - "hide": 0, - "includeAll": false, - "label": "Name", - "multi": false, - "name": "statefulset", - "options": [ - - ], - "query": "label_values(kube_statefulset_metadata_generation{job=\"kube-state-metrics\", cluster=\"$cluster\", namespace=\"$namespace\"}, statefulset)", - "refresh": 2, - "regex": "", - "sort": 1, - "tagValuesQuery": "", - "tags": [ - - ], - "tagsQuery": "", - "type": "query", - "useTags": false - } - ] - }, - "time": { - "from": "now-1h", - "to": "now" - }, - "timepicker": { - "refresh_intervals": [ - "5s", - "10s", - "30s", - "1m", - "5m", - "15m", - "30m", - "1h", - "2h", - "1d" - ], - "time_options": [ - "5m", - "15m", - "1h", - "6h", - "12h", - "24h", - "2d", - "7d", - "30d" - ] - }, - "timezone": "UTC", - "title": "Kubernetes / StatefulSets", - "uid": "a31c1f46e6f727cb37c0d731a7245005", - "version": 0 - } kind: ConfigMap metadata: name: grafana-dashboards-k8s diff --git a/addons/grafana/dashboards-node-exporter.yaml b/addons/grafana/dashboards-node-exporter.yaml index d5aac1bf..8740d532 100644 --- a/addons/grafana/dashboards-node-exporter.yaml +++ b/addons/grafana/dashboards-node-exporter.yaml @@ -15,13 +15,13 @@ data: }, "editable": false, "gnetId": null, - "graphTooltip": 0, + "graphTooltip": 1, "hideControls": false, "id": null, "links": [ ], - "refresh": "", + "refresh": "30s", "rows": [ { "collapse": false, @@ -73,9 +73,8 @@ data: "steppedLine": false, "targets": [ { - "expr": "(\n (1 - rate(node_cpu_seconds_total{job=\"node-exporter\", mode=\"idle\", instance=\"$instance\"}[$__interval]))\n/ ignoring(cpu) group_left\n count without (cpu)( node_cpu_seconds_total{job=\"node-exporter\", mode=\"idle\", instance=\"$instance\"})\n)\n", + "expr": "(\n (1 - sum without (mode) (rate(node_cpu_seconds_total{job=\"node-exporter\", mode=~\"idle|iowait|steal\", instance=\"$instance\"}[$__rate_interval])))\n/ ignoring(cpu) group_left\n count without (cpu, mode) (node_cpu_seconds_total{job=\"node-exporter\", mode=\"idle\", instance=\"$instance\"})\n)\n", "format": "time_series", - "interval": "1m", "intervalFactor": 5, "legendFormat": "{{cpu}}", "refId": "A" @@ -509,25 +508,22 @@ data: "steppedLine": false, "targets": [ { - "expr": "rate(node_disk_read_bytes_total{job=\"node-exporter\", instance=\"$instance\", device!~\"dm.*\"}[$__interval])", + "expr": "rate(node_disk_read_bytes_total{job=\"node-exporter\", instance=\"$instance\", device!~\"dm.*\"}[$__rate_interval])", "format": "time_series", - "interval": "1m", "intervalFactor": 2, "legendFormat": "{{device}} read", "refId": "A" }, { - "expr": "rate(node_disk_written_bytes_total{job=\"node-exporter\", instance=\"$instance\", device!~\"dm.*\"}[$__interval])", + "expr": "rate(node_disk_written_bytes_total{job=\"node-exporter\", instance=\"$instance\", device!~\"dm.*\"}[$__rate_interval])", "format": "time_series", - "interval": "1m", "intervalFactor": 2, "legendFormat": "{{device}} written", "refId": "B" }, { - "expr": "rate(node_disk_io_time_seconds_total{job=\"node-exporter\", instance=\"$instance\", device!~\"dm.*\"}[$__interval])", + "expr": "rate(node_disk_io_time_seconds_total{job=\"node-exporter\", instance=\"$instance\", device!~\"dm.*\"}[$__rate_interval])", "format": "time_series", - "interval": "1m", "intervalFactor": 2, "legendFormat": "{{device}} io time", "refId": "C" @@ -739,9 +735,8 @@ data: "steppedLine": false, "targets": [ { - "expr": "rate(node_network_receive_bytes_total{job=\"node-exporter\", instance=\"$instance\", device!=\"lo\"}[$__interval])", + "expr": "rate(node_network_receive_bytes_total{job=\"node-exporter\", instance=\"$instance\", device!=\"lo\"}[$__rate_interval])", "format": "time_series", - "interval": "1m", "intervalFactor": 2, "legendFormat": "{{device}}", "refId": "A" @@ -833,9 +828,8 @@ data: "steppedLine": false, "targets": [ { - "expr": "rate(node_network_transmit_bytes_total{job=\"node-exporter\", instance=\"$instance\", device!=\"lo\"}[$__interval])", + "expr": "rate(node_network_transmit_bytes_total{job=\"node-exporter\", instance=\"$instance\", device!=\"lo\"}[$__rate_interval])", "format": "time_series", - "interval": "1m", "intervalFactor": 2, "legendFormat": "{{device}}", "refId": "A" @@ -894,17 +888,17 @@ data: "schemaVersion": 14, "style": "dark", "tags": [ - + "node-exporter-mixin" ], "templating": { "list": [ { "current": { - "text": "Prometheus", - "value": "Prometheus" + "text": "default", + "value": "default" }, "hide": 0, - "label": null, + "label": "Data Source", "name": "datasource", "options": [ @@ -972,7 +966,7 @@ data: ] }, "timezone": "", - "title": "Nodes", + "title": "Node Exporter / Nodes", "uid": "fa49a4706d07a042595b664c87fb33ea", "version": 0 } diff --git a/addons/grafana/dashboards-prom.yaml b/addons/grafana/dashboards-prom.yaml index 94638cb4..155e8a7e 100644 --- a/addons/grafana/dashboards-prom.yaml +++ b/addons/grafana/dashboards-prom.yaml @@ -1536,11 +1536,11 @@ data: "includeAll": true, "label": null, "multi": false, - "name": "instance", + "name": "cluster", "options": [ ], - "query": "label_values(prometheus_build_info, instance)", + "query": "label_values(kube_pod_container_info{image=~\".*prometheus.*\"}, cluster)", "refresh": 2, "regex": "", "sort": 0, @@ -1571,11 +1571,11 @@ data: "includeAll": true, "label": null, "multi": false, - "name": "cluster", + "name": "instance", "options": [ ], - "query": "label_values(kube_pod_container_info{image=~\".*prometheus.*\"}, cluster)", + "query": "label_values(prometheus_build_info{cluster=~\"$cluster\"}, instance)", "refresh": 2, "regex": "", "sort": 0, @@ -1850,7 +1850,7 @@ data: "title": "Prometheus Stats", "tooltip": { "shared": true, - "sort": 0, + "sort": 2, "value_type": "individual" }, "transform": "table", @@ -1949,7 +1949,7 @@ data: "title": "Target Sync", "tooltip": { "shared": true, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -2035,7 +2035,7 @@ data: "title": "Targets", "tooltip": { "shared": true, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -2133,7 +2133,7 @@ data: "title": "Average Scrape Interval Duration", "tooltip": { "shared": true, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -2202,6 +2202,14 @@ data: "stack": true, "steppedLine": false, "targets": [ + { + "expr": "sum by (job) (rate(prometheus_target_scrapes_exceeded_body_size_limit_total[1m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "exceeded body size limit: {{job}}", + "legendLink": null, + "step": 10 + }, { "expr": "sum by (job) (rate(prometheus_target_scrapes_exceeded_sample_limit_total[1m]))", "format": "time_series", @@ -2243,7 +2251,7 @@ data: "title": "Scrape failures", "tooltip": { "shared": true, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -2329,7 +2337,7 @@ data: "title": "Appended Samples", "tooltip": { "shared": true, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -2427,7 +2435,7 @@ data: "title": "Head Series", "tooltip": { "shared": true, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -2513,7 +2521,7 @@ data: "title": "Head Chunks", "tooltip": { "shared": true, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -2611,7 +2619,7 @@ data: "title": "Query Rate", "tooltip": { "shared": true, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -2697,7 +2705,7 @@ data: "title": "Stage Duration", "tooltip": { "shared": true, - "sort": 0, + "sort": 2, "value_type": "individual" }, "type": "graph", @@ -2751,7 +2759,7 @@ data: "value": "default" }, "hide": 0, - "label": null, + "label": "Data Source", "name": "datasource", "options": [ @@ -2762,7 +2770,7 @@ data: "type": "datasource" }, { - "allValue": null, + "allValue": ".+", "current": { "selected": true, "text": "All", @@ -2777,7 +2785,7 @@ data: "options": [ ], - "query": "label_values(prometheus_build_info, job)", + "query": "label_values(prometheus_build_info{job=\"prometheus\"}, job)", "refresh": 1, "regex": "", "sort": 2, @@ -2790,7 +2798,7 @@ data: "useTags": false }, { - "allValue": null, + "allValue": ".+", "current": { "selected": true, "text": "All", @@ -2805,7 +2813,7 @@ data: "options": [ ], - "query": "label_values(prometheus_build_info, instance)", + "query": "label_values(prometheus_build_info{job=~\"$job\"}, instance)", "refresh": 1, "regex": "", "sort": 2, diff --git a/addons/grafana/deployment.yaml b/addons/grafana/deployment.yaml index 36b29596..df079280 100644 --- a/addons/grafana/deployment.yaml +++ b/addons/grafana/deployment.yaml @@ -69,6 +69,8 @@ spec: mountPath: /etc/grafana/dashboards/k8s-resources-1 - name: dashboards-k8s-resources-2 mountPath: /etc/grafana/dashboards/k8s-resources-2 + - name: dashboards-k8s-network + mountPath: /etc/grafana/dashboards/k8s-network - name: dashboards-coredns mountPath: /etc/grafana/dashboards/coredns - name: dashboards-nginx-ingress @@ -101,6 +103,9 @@ spec: - name: dashboards-k8s-resources-1 configMap: name: grafana-dashboards-k8s-resources-1 + - name: dashboards-k8s-network + configMap: + name: grafana-dashboards-k8s-network - name: dashboards-k8s-resources-2 configMap: name: grafana-dashboards-k8s-resources-2 diff --git a/addons/prometheus/rules.yaml b/addons/prometheus/rules.yaml index 51504668..98aa5bbf 100644 --- a/addons/prometheus/rules.yaml +++ b/addons/prometheus/rules.yaml @@ -57,10 +57,10 @@ data: { "alert": "etcdGRPCRequestsSlow", "annotations": { - "description": "etcd cluster \"{{ $labels.job }}\": gRPC requests to {{ $labels.grpc_method }} are taking {{ $value }}s on etcd instance {{ $labels.instance }}.", + "description": "etcd cluster \"{{ $labels.job }}\": 99th percentile of gRPC requests is {{ $value }}s on etcd instance {{ $labels.instance }} for {{ $labels.grpc_method }} method.", "summary": "etcd grpc requests are slow" }, - "expr": "histogram_quantile(0.99, sum(rate(grpc_server_handling_seconds_bucket{job=~\".*etcd.*\", grpc_type=\"unary\"}[5m])) without(grpc_type))\n> 0.15\n", + "expr": "histogram_quantile(0.99, sum(rate(grpc_server_handling_seconds_bucket{job=~\".*etcd.*\", grpc_method!=\"Defragment\", grpc_type=\"unary\"}[5m])) without(grpc_type))\n> 0.15\n", "for": "10m", "labels": { "severity": "critical" @@ -105,7 +105,8 @@ data: { "alert": "etcdHighFsyncDurations", "annotations": { - "message": "etcd cluster \"{{ $labels.job }}\": 99th percentile fsync durations are {{ $value }}s on etcd instance {{ $labels.instance }}." + "description": "etcd cluster \"{{ $labels.job }}\": 99th percentile fsync durations are {{ $value }}s on etcd instance {{ $labels.instance }}.", + "summary": "etcd cluster 99th percentile fsync durations are too high." }, "expr": "histogram_quantile(0.99, rate(etcd_disk_wal_fsync_duration_seconds_bucket{job=~\".*etcd.*\"}[5m]))\n> 1\n", "for": "10m", @@ -125,46 +126,11 @@ data: "severity": "warning" } }, - { - "alert": "etcdHighNumberOfFailedHTTPRequests", - "annotations": { - "description": "{{ $value }}% of requests for {{ $labels.method }} failed on etcd instance {{ $labels.instance }}", - "summary": "etcd has high number of failed HTTP requests." - }, - "expr": "sum(rate(etcd_http_failed_total{job=~\".*etcd.*\", code!=\"404\"}[5m])) without (code) / sum(rate(etcd_http_received_total{job=~\".*etcd.*\"}[5m]))\nwithout (code) > 0.01\n", - "for": "10m", - "labels": { - "severity": "warning" - } - }, - { - "alert": "etcdHighNumberOfFailedHTTPRequests", - "annotations": { - "description": "{{ $value }}% of requests for {{ $labels.method }} failed on etcd instance {{ $labels.instance }}.", - "summary": "etcd has high number of failed HTTP requests." - }, - "expr": "sum(rate(etcd_http_failed_total{job=~\".*etcd.*\", code!=\"404\"}[5m])) without (code) / sum(rate(etcd_http_received_total{job=~\".*etcd.*\"}[5m]))\nwithout (code) > 0.05\n", - "for": "10m", - "labels": { - "severity": "critical" - } - }, - { - "alert": "etcdHTTPRequestsSlow", - "annotations": { - "description": "etcd instance {{ $labels.instance }} HTTP requests to {{ $labels.method }} are slow.", - "summary": "etcd instance HTTP requests are slow." - }, - "expr": "histogram_quantile(0.99, rate(etcd_http_successful_duration_seconds_bucket[5m]))\n> 0.15\n", - "for": "10m", - "labels": { - "severity": "warning" - } - }, { "alert": "etcdBackendQuotaLowSpace", "annotations": { - "message": "etcd cluster \"{{ $labels.job }}\": database size exceeds the defined quota on etcd instance {{ $labels.instance }}, please defrag or increase the quota as the writes to etcd will be disabled when it is full." + "description": "etcd cluster \"{{ $labels.job }}\": database size exceeds the defined quota on etcd instance {{ $labels.instance }}, please defrag or increase the quota as the writes to etcd will be disabled when it is full.", + "summary": "etcd cluster database is running full." }, "expr": "(etcd_mvcc_db_total_size_in_bytes/etcd_server_quota_backend_bytes)*100 > 95\n", "for": "10m", @@ -175,7 +141,8 @@ data: { "alert": "etcdExcessiveDatabaseGrowth", "annotations": { - "message": "etcd cluster \"{{ $labels.job }}\": Observed surge in etcd writes leading to 50% increase in database size over the past four hours on etcd instance {{ $labels.instance }}, please check as it might be disruptive." + "description": "etcd cluster \"{{ $labels.job }}\": Observed surge in etcd writes leading to 50% increase in database size over the past four hours on etcd instance {{ $labels.instance }}, please check as it might be disruptive.", + "summary": "etcd cluster database growing very fast." }, "expr": "increase(((etcd_mvcc_db_total_size_in_bytes/etcd_server_quota_backend_bytes)*100)[240m:1m]) > 50\n", "for": "10m", @@ -191,122 +158,113 @@ data: { "groups": [ { - "name": "kube-apiserver.rules", + "name": "kube-apiserver-burnrate.rules", "rules": [ { - "expr": "(\n (\n # too slow\n sum(rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"LIST|GET\"}[1d]))\n -\n (\n (\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=~\"resource|\",le=\"0.1\"}[1d]))\n or\n vector(0)\n )\n +\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=\"namespace\",le=\"0.5\"}[1d]))\n +\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=\"cluster\",le=\"5\"}[1d]))\n )\n )\n +\n # errors\n sum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\",code=~\"5..\"}[1d]))\n)\n/\nsum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\"}[1d]))\n", + "expr": "(\n (\n # too slow\n sum by (cluster) (rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\"}[1d]))\n -\n (\n (\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=~\"resource|\",le=\"1\"}[1d]))\n or\n vector(0)\n )\n +\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=\"namespace\",le=\"5\"}[1d]))\n +\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=\"cluster\",le=\"30\"}[1d]))\n )\n )\n +\n # errors\n sum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\",code=~\"5..\"}[1d]))\n)\n/\nsum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\"}[1d]))\n", "labels": { "verb": "read" }, "record": "apiserver_request:burnrate1d" }, { - "expr": "(\n (\n # too slow\n sum(rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"LIST|GET\"}[1h]))\n -\n (\n (\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=~\"resource|\",le=\"0.1\"}[1h]))\n or\n vector(0)\n )\n +\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=\"namespace\",le=\"0.5\"}[1h]))\n +\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=\"cluster\",le=\"5\"}[1h]))\n )\n )\n +\n # errors\n sum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\",code=~\"5..\"}[1h]))\n)\n/\nsum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\"}[1h]))\n", + "expr": "(\n (\n # too slow\n sum by (cluster) (rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\"}[1h]))\n -\n (\n (\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=~\"resource|\",le=\"1\"}[1h]))\n or\n vector(0)\n )\n +\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=\"namespace\",le=\"5\"}[1h]))\n +\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=\"cluster\",le=\"30\"}[1h]))\n )\n )\n +\n # errors\n sum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\",code=~\"5..\"}[1h]))\n)\n/\nsum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\"}[1h]))\n", "labels": { "verb": "read" }, "record": "apiserver_request:burnrate1h" }, { - "expr": "(\n (\n # too slow\n sum(rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"LIST|GET\"}[2h]))\n -\n (\n (\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=~\"resource|\",le=\"0.1\"}[2h]))\n or\n vector(0)\n )\n +\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=\"namespace\",le=\"0.5\"}[2h]))\n +\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=\"cluster\",le=\"5\"}[2h]))\n )\n )\n +\n # errors\n sum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\",code=~\"5..\"}[2h]))\n)\n/\nsum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\"}[2h]))\n", + "expr": "(\n (\n # too slow\n sum by (cluster) (rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\"}[2h]))\n -\n (\n (\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=~\"resource|\",le=\"1\"}[2h]))\n or\n vector(0)\n )\n +\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=\"namespace\",le=\"5\"}[2h]))\n +\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=\"cluster\",le=\"30\"}[2h]))\n )\n )\n +\n # errors\n sum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\",code=~\"5..\"}[2h]))\n)\n/\nsum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\"}[2h]))\n", "labels": { "verb": "read" }, "record": "apiserver_request:burnrate2h" }, { - "expr": "(\n (\n # too slow\n sum(rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"LIST|GET\"}[30m]))\n -\n (\n (\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=~\"resource|\",le=\"0.1\"}[30m]))\n or\n vector(0)\n )\n +\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=\"namespace\",le=\"0.5\"}[30m]))\n +\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=\"cluster\",le=\"5\"}[30m]))\n )\n )\n +\n # errors\n sum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\",code=~\"5..\"}[30m]))\n)\n/\nsum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\"}[30m]))\n", + "expr": "(\n (\n # too slow\n sum by (cluster) (rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\"}[30m]))\n -\n (\n (\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=~\"resource|\",le=\"1\"}[30m]))\n or\n vector(0)\n )\n +\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=\"namespace\",le=\"5\"}[30m]))\n +\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=\"cluster\",le=\"30\"}[30m]))\n )\n )\n +\n # errors\n sum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\",code=~\"5..\"}[30m]))\n)\n/\nsum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\"}[30m]))\n", "labels": { "verb": "read" }, "record": "apiserver_request:burnrate30m" }, { - "expr": "(\n (\n # too slow\n sum(rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"LIST|GET\"}[3d]))\n -\n (\n (\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=~\"resource|\",le=\"0.1\"}[3d]))\n or\n vector(0)\n )\n +\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=\"namespace\",le=\"0.5\"}[3d]))\n +\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=\"cluster\",le=\"5\"}[3d]))\n )\n )\n +\n # errors\n sum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\",code=~\"5..\"}[3d]))\n)\n/\nsum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\"}[3d]))\n", + "expr": "(\n (\n # too slow\n sum by (cluster) (rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\"}[3d]))\n -\n (\n (\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=~\"resource|\",le=\"1\"}[3d]))\n or\n vector(0)\n )\n +\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=\"namespace\",le=\"5\"}[3d]))\n +\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=\"cluster\",le=\"30\"}[3d]))\n )\n )\n +\n # errors\n sum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\",code=~\"5..\"}[3d]))\n)\n/\nsum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\"}[3d]))\n", "labels": { "verb": "read" }, "record": "apiserver_request:burnrate3d" }, { - "expr": "(\n (\n # too slow\n sum(rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"LIST|GET\"}[5m]))\n -\n (\n (\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=~\"resource|\",le=\"0.1\"}[5m]))\n or\n vector(0)\n )\n +\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=\"namespace\",le=\"0.5\"}[5m]))\n +\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=\"cluster\",le=\"5\"}[5m]))\n )\n )\n +\n # errors\n sum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\",code=~\"5..\"}[5m]))\n)\n/\nsum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\"}[5m]))\n", + "expr": "(\n (\n # too slow\n sum by (cluster) (rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\"}[5m]))\n -\n (\n (\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=~\"resource|\",le=\"1\"}[5m]))\n or\n vector(0)\n )\n +\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=\"namespace\",le=\"5\"}[5m]))\n +\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=\"cluster\",le=\"30\"}[5m]))\n )\n )\n +\n # errors\n sum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\",code=~\"5..\"}[5m]))\n)\n/\nsum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\"}[5m]))\n", "labels": { "verb": "read" }, "record": "apiserver_request:burnrate5m" }, { - "expr": "(\n (\n # too slow\n sum(rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"LIST|GET\"}[6h]))\n -\n (\n (\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=~\"resource|\",le=\"0.1\"}[6h]))\n or\n vector(0)\n )\n +\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=\"namespace\",le=\"0.5\"}[6h]))\n +\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=\"cluster\",le=\"5\"}[6h]))\n )\n )\n +\n # errors\n sum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\",code=~\"5..\"}[6h]))\n)\n/\nsum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\"}[6h]))\n", + "expr": "(\n (\n # too slow\n sum by (cluster) (rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\"}[6h]))\n -\n (\n (\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=~\"resource|\",le=\"1\"}[6h]))\n or\n vector(0)\n )\n +\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=\"namespace\",le=\"5\"}[6h]))\n +\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\",scope=\"cluster\",le=\"30\"}[6h]))\n )\n )\n +\n # errors\n sum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\",code=~\"5..\"}[6h]))\n)\n/\nsum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\"}[6h]))\n", "labels": { "verb": "read" }, "record": "apiserver_request:burnrate6h" }, { - "expr": "(\n (\n # too slow\n sum(rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[1d]))\n -\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",le=\"1\"}[1d]))\n )\n +\n sum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",code=~\"5..\"}[1d]))\n)\n/\nsum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[1d]))\n", + "expr": "(\n (\n # too slow\n sum by (cluster) (rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",subresource!~\"proxy|attach|log|exec|portforward\"}[1d]))\n -\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",subresource!~\"proxy|attach|log|exec|portforward\",le=\"1\"}[1d]))\n )\n +\n sum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",code=~\"5..\"}[1d]))\n)\n/\nsum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[1d]))\n", "labels": { "verb": "write" }, "record": "apiserver_request:burnrate1d" }, { - "expr": "(\n (\n # too slow\n sum(rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[1h]))\n -\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",le=\"1\"}[1h]))\n )\n +\n sum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",code=~\"5..\"}[1h]))\n)\n/\nsum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[1h]))\n", + "expr": "(\n (\n # too slow\n sum by (cluster) (rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",subresource!~\"proxy|attach|log|exec|portforward\"}[1h]))\n -\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",subresource!~\"proxy|attach|log|exec|portforward\",le=\"1\"}[1h]))\n )\n +\n sum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",code=~\"5..\"}[1h]))\n)\n/\nsum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[1h]))\n", "labels": { "verb": "write" }, "record": "apiserver_request:burnrate1h" }, { - "expr": "(\n (\n # too slow\n sum(rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[2h]))\n -\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",le=\"1\"}[2h]))\n )\n +\n sum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",code=~\"5..\"}[2h]))\n)\n/\nsum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[2h]))\n", + "expr": "(\n (\n # too slow\n sum by (cluster) (rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",subresource!~\"proxy|attach|log|exec|portforward\"}[2h]))\n -\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",subresource!~\"proxy|attach|log|exec|portforward\",le=\"1\"}[2h]))\n )\n +\n sum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",code=~\"5..\"}[2h]))\n)\n/\nsum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[2h]))\n", "labels": { "verb": "write" }, "record": "apiserver_request:burnrate2h" }, { - "expr": "(\n (\n # too slow\n sum(rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[30m]))\n -\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",le=\"1\"}[30m]))\n )\n +\n sum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",code=~\"5..\"}[30m]))\n)\n/\nsum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[30m]))\n", + "expr": "(\n (\n # too slow\n sum by (cluster) (rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",subresource!~\"proxy|attach|log|exec|portforward\"}[30m]))\n -\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",subresource!~\"proxy|attach|log|exec|portforward\",le=\"1\"}[30m]))\n )\n +\n sum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",code=~\"5..\"}[30m]))\n)\n/\nsum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[30m]))\n", "labels": { "verb": "write" }, "record": "apiserver_request:burnrate30m" }, { - "expr": "(\n (\n # too slow\n sum(rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[3d]))\n -\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",le=\"1\"}[3d]))\n )\n +\n sum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",code=~\"5..\"}[3d]))\n)\n/\nsum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[3d]))\n", + "expr": "(\n (\n # too slow\n sum by (cluster) (rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",subresource!~\"proxy|attach|log|exec|portforward\"}[3d]))\n -\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",subresource!~\"proxy|attach|log|exec|portforward\",le=\"1\"}[3d]))\n )\n +\n sum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",code=~\"5..\"}[3d]))\n)\n/\nsum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[3d]))\n", "labels": { "verb": "write" }, "record": "apiserver_request:burnrate3d" }, { - "expr": "(\n (\n # too slow\n sum(rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[5m]))\n -\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",le=\"1\"}[5m]))\n )\n +\n sum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",code=~\"5..\"}[5m]))\n)\n/\nsum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[5m]))\n", + "expr": "(\n (\n # too slow\n sum by (cluster) (rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",subresource!~\"proxy|attach|log|exec|portforward\"}[5m]))\n -\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",subresource!~\"proxy|attach|log|exec|portforward\",le=\"1\"}[5m]))\n )\n +\n sum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",code=~\"5..\"}[5m]))\n)\n/\nsum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[5m]))\n", "labels": { "verb": "write" }, "record": "apiserver_request:burnrate5m" }, { - "expr": "(\n (\n # too slow\n sum(rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[6h]))\n -\n sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",le=\"1\"}[6h]))\n )\n +\n sum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",code=~\"5..\"}[6h]))\n)\n/\nsum(rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[6h]))\n", + "expr": "(\n (\n # too slow\n sum by (cluster) (rate(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",subresource!~\"proxy|attach|log|exec|portforward\"}[6h]))\n -\n sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",subresource!~\"proxy|attach|log|exec|portforward\",le=\"1\"}[6h]))\n )\n +\n sum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",code=~\"5..\"}[6h]))\n)\n/\nsum by (cluster) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[6h]))\n", "labels": { "verb": "write" }, "record": "apiserver_request:burnrate6h" - }, + } + ] + }, + { + "name": "kube-apiserver-histogram.rules", + "rules": [ { - "expr": "sum by (code,resource) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\"}[5m]))\n", - "labels": { - "verb": "read" - }, - "record": "code_resource:apiserver_request_total:rate5m" - }, - { - "expr": "sum by (code,resource) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[5m]))\n", - "labels": { - "verb": "write" - }, - "record": "code_resource:apiserver_request_total:rate5m" - }, - { - "expr": "histogram_quantile(0.99, sum by (le, resource) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\"}[5m]))) > 0\n", + "expr": "histogram_quantile(0.99, sum by (cluster, le, resource) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",subresource!~\"proxy|attach|log|exec|portforward\"}[5m]))) > 0\n", "labels": { "quantile": "0.99", "verb": "read" @@ -314,33 +272,12 @@ data: "record": "cluster_quantile:apiserver_request_duration_seconds:histogram_quantile" }, { - "expr": "histogram_quantile(0.99, sum by (le, resource) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[5m]))) > 0\n", + "expr": "histogram_quantile(0.99, sum by (cluster, le, resource) (rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\",subresource!~\"proxy|attach|log|exec|portforward\"}[5m]))) > 0\n", "labels": { "quantile": "0.99", "verb": "write" }, "record": "cluster_quantile:apiserver_request_duration_seconds:histogram_quantile" - }, - { - "expr": "histogram_quantile(0.99, sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",subresource!=\"log\",verb!~\"LIST|WATCH|WATCHLIST|DELETECOLLECTION|PROXY|CONNECT\"}[5m])) without(instance, pod))\n", - "labels": { - "quantile": "0.99" - }, - "record": "cluster_quantile:apiserver_request_duration_seconds:histogram_quantile" - }, - { - "expr": "histogram_quantile(0.9, sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",subresource!=\"log\",verb!~\"LIST|WATCH|WATCHLIST|DELETECOLLECTION|PROXY|CONNECT\"}[5m])) without(instance, pod))\n", - "labels": { - "quantile": "0.9" - }, - "record": "cluster_quantile:apiserver_request_duration_seconds:histogram_quantile" - }, - { - "expr": "histogram_quantile(0.5, sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\",subresource!=\"log\",verb!~\"LIST|WATCH|WATCHLIST|DELETECOLLECTION|PROXY|CONNECT\"}[5m])) without(instance, pod))\n", - "labels": { - "quantile": "0.5" - }, - "record": "cluster_quantile:apiserver_request_duration_seconds:histogram_quantile" } ] }, @@ -349,135 +286,89 @@ data: "name": "kube-apiserver-availability.rules", "rules": [ { - "expr": "1 - (\n (\n # write too slow\n sum(increase(apiserver_request_duration_seconds_count{verb=~\"POST|PUT|PATCH|DELETE\"}[30d]))\n -\n sum(increase(apiserver_request_duration_seconds_bucket{verb=~\"POST|PUT|PATCH|DELETE\",le=\"1\"}[30d]))\n ) +\n (\n # read too slow\n sum(increase(apiserver_request_duration_seconds_count{verb=~\"LIST|GET\"}[30d]))\n -\n (\n (\n sum(increase(apiserver_request_duration_seconds_bucket{verb=~\"LIST|GET\",scope=~\"resource|\",le=\"0.1\"}[30d]))\n or\n vector(0)\n )\n +\n sum(increase(apiserver_request_duration_seconds_bucket{verb=~\"LIST|GET\",scope=\"namespace\",le=\"0.5\"}[30d]))\n +\n sum(increase(apiserver_request_duration_seconds_bucket{verb=~\"LIST|GET\",scope=\"cluster\",le=\"5\"}[30d]))\n )\n ) +\n # errors\n sum(code:apiserver_request_total:increase30d{code=~\"5..\"} or vector(0))\n)\n/\nsum(code:apiserver_request_total:increase30d)\n", + "expr": "avg_over_time(code_verb:apiserver_request_total:increase1h[30d]) * 24 * 30\n", + "record": "code_verb:apiserver_request_total:increase30d" + }, + { + "expr": "sum by (cluster, code) (code_verb:apiserver_request_total:increase30d{verb=~\"LIST|GET\"})\n", + "labels": { + "verb": "read" + }, + "record": "code:apiserver_request_total:increase30d" + }, + { + "expr": "sum by (cluster, code) (code_verb:apiserver_request_total:increase30d{verb=~\"POST|PUT|PATCH|DELETE\"})\n", + "labels": { + "verb": "write" + }, + "record": "code:apiserver_request_total:increase30d" + }, + { + "expr": "sum by (cluster, verb, scope) (increase(apiserver_request_duration_seconds_count[1h]))\n", + "record": "cluster_verb_scope:apiserver_request_duration_seconds_count:increase1h" + }, + { + "expr": "sum by (cluster, verb, scope) (avg_over_time(cluster_verb_scope:apiserver_request_duration_seconds_count:increase1h[30d]) * 24 * 30)\n", + "record": "cluster_verb_scope:apiserver_request_duration_seconds_count:increase30d" + }, + { + "expr": "sum by (cluster, verb, scope, le) (increase(apiserver_request_duration_seconds_bucket[1h]))\n", + "record": "cluster_verb_scope_le:apiserver_request_duration_seconds_bucket:increase1h" + }, + { + "expr": "sum by (cluster, verb, scope, le) (avg_over_time(cluster_verb_scope_le:apiserver_request_duration_seconds_bucket:increase1h[30d]) * 24 * 30)\n", + "record": "cluster_verb_scope_le:apiserver_request_duration_seconds_bucket:increase30d" + }, + { + "expr": "1 - (\n (\n # write too slow\n sum by (cluster) (cluster_verb_scope:apiserver_request_duration_seconds_count:increase30d{verb=~\"POST|PUT|PATCH|DELETE\"})\n -\n sum by (cluster) (cluster_verb_scope_le:apiserver_request_duration_seconds_bucket:increase30d{verb=~\"POST|PUT|PATCH|DELETE\",le=\"1\"})\n ) +\n (\n # read too slow\n sum by (cluster) (cluster_verb_scope:apiserver_request_duration_seconds_count:increase30d{verb=~\"LIST|GET\"})\n -\n (\n (\n sum by (cluster) (cluster_verb_scope_le:apiserver_request_duration_seconds_bucket:increase30d{verb=~\"LIST|GET\",scope=~\"resource|\",le=\"1\"})\n or\n vector(0)\n )\n +\n sum by (cluster) (cluster_verb_scope_le:apiserver_request_duration_seconds_bucket:increase30d{verb=~\"LIST|GET\",scope=\"namespace\",le=\"5\"})\n +\n sum by (cluster) (cluster_verb_scope_le:apiserver_request_duration_seconds_bucket:increase30d{verb=~\"LIST|GET\",scope=\"cluster\",le=\"30\"})\n )\n ) +\n # errors\n sum by (cluster) (code:apiserver_request_total:increase30d{code=~\"5..\"} or vector(0))\n)\n/\nsum by (cluster) (code:apiserver_request_total:increase30d)\n", "labels": { "verb": "all" }, "record": "apiserver_request:availability30d" }, { - "expr": "1 - (\n sum(increase(apiserver_request_duration_seconds_count{job=\"apiserver\",verb=~\"LIST|GET\"}[30d]))\n -\n (\n # too slow\n (\n sum(increase(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=~\"resource|\",le=\"0.1\"}[30d]))\n or\n vector(0)\n )\n +\n sum(increase(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=\"namespace\",le=\"0.5\"}[30d]))\n +\n sum(increase(apiserver_request_duration_seconds_bucket{job=\"apiserver\",verb=~\"LIST|GET\",scope=\"cluster\",le=\"5\"}[30d]))\n )\n +\n # errors\n sum(code:apiserver_request_total:increase30d{verb=\"read\",code=~\"5..\"} or vector(0))\n)\n/\nsum(code:apiserver_request_total:increase30d{verb=\"read\"})\n", + "expr": "1 - (\n sum by (cluster) (cluster_verb_scope:apiserver_request_duration_seconds_count:increase30d{verb=~\"LIST|GET\"})\n -\n (\n # too slow\n (\n sum by (cluster) (cluster_verb_scope_le:apiserver_request_duration_seconds_bucket:increase30d{verb=~\"LIST|GET\",scope=~\"resource|\",le=\"1\"})\n or\n vector(0)\n )\n +\n sum by (cluster) (cluster_verb_scope_le:apiserver_request_duration_seconds_bucket:increase30d{verb=~\"LIST|GET\",scope=\"namespace\",le=\"5\"})\n +\n sum by (cluster) (cluster_verb_scope_le:apiserver_request_duration_seconds_bucket:increase30d{verb=~\"LIST|GET\",scope=\"cluster\",le=\"30\"})\n )\n +\n # errors\n sum by (cluster) (code:apiserver_request_total:increase30d{verb=\"read\",code=~\"5..\"} or vector(0))\n)\n/\nsum by (cluster) (code:apiserver_request_total:increase30d{verb=\"read\"})\n", "labels": { "verb": "read" }, "record": "apiserver_request:availability30d" }, { - "expr": "1 - (\n (\n # too slow\n sum(increase(apiserver_request_duration_seconds_count{verb=~\"POST|PUT|PATCH|DELETE\"}[30d]))\n -\n sum(increase(apiserver_request_duration_seconds_bucket{verb=~\"POST|PUT|PATCH|DELETE\",le=\"1\"}[30d]))\n )\n +\n # errors\n sum(code:apiserver_request_total:increase30d{verb=\"write\",code=~\"5..\"} or vector(0))\n)\n/\nsum(code:apiserver_request_total:increase30d{verb=\"write\"})\n", + "expr": "1 - (\n (\n # too slow\n sum by (cluster) (cluster_verb_scope:apiserver_request_duration_seconds_count:increase30d{verb=~\"POST|PUT|PATCH|DELETE\"})\n -\n sum by (cluster) (cluster_verb_scope_le:apiserver_request_duration_seconds_bucket:increase30d{verb=~\"POST|PUT|PATCH|DELETE\",le=\"1\"})\n )\n +\n # errors\n sum by (cluster) (code:apiserver_request_total:increase30d{verb=\"write\",code=~\"5..\"} or vector(0))\n)\n/\nsum by (cluster) (code:apiserver_request_total:increase30d{verb=\"write\"})\n", "labels": { "verb": "write" }, "record": "apiserver_request:availability30d" }, { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"LIST\",code=~\"2..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"GET\",code=~\"2..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"POST\",code=~\"2..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"PUT\",code=~\"2..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"PATCH\",code=~\"2..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"DELETE\",code=~\"2..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"LIST\",code=~\"3..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"GET\",code=~\"3..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"POST\",code=~\"3..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"PUT\",code=~\"3..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"PATCH\",code=~\"3..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"DELETE\",code=~\"3..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"LIST\",code=~\"4..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"GET\",code=~\"4..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"POST\",code=~\"4..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"PUT\",code=~\"4..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"PATCH\",code=~\"4..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"DELETE\",code=~\"4..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"LIST\",code=~\"5..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"GET\",code=~\"5..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"POST\",code=~\"5..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"PUT\",code=~\"5..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"PATCH\",code=~\"5..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=\"DELETE\",code=~\"5..\"}[30d]))\n", - "record": "code_verb:apiserver_request_total:increase30d" - }, - { - "expr": "sum by (code) (code_verb:apiserver_request_total:increase30d{verb=~\"LIST|GET\"})\n", + "expr": "sum by (cluster,code,resource) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET\"}[5m]))\n", "labels": { "verb": "read" }, - "record": "code:apiserver_request_total:increase30d" + "record": "code_resource:apiserver_request_total:rate5m" }, { - "expr": "sum by (code) (code_verb:apiserver_request_total:increase30d{verb=~\"POST|PUT|PATCH|DELETE\"})\n", + "expr": "sum by (cluster,code,resource) (rate(apiserver_request_total{job=\"apiserver\",verb=~\"POST|PUT|PATCH|DELETE\"}[5m]))\n", "labels": { "verb": "write" }, - "record": "code:apiserver_request_total:increase30d" + "record": "code_resource:apiserver_request_total:rate5m" + }, + { + "expr": "sum by (cluster, code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET|POST|PUT|PATCH|DELETE\",code=~\"2..\"}[1h]))\n", + "record": "code_verb:apiserver_request_total:increase1h" + }, + { + "expr": "sum by (cluster, code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET|POST|PUT|PATCH|DELETE\",code=~\"3..\"}[1h]))\n", + "record": "code_verb:apiserver_request_total:increase1h" + }, + { + "expr": "sum by (cluster, code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET|POST|PUT|PATCH|DELETE\",code=~\"4..\"}[1h]))\n", + "record": "code_verb:apiserver_request_total:increase1h" + }, + { + "expr": "sum by (cluster, code, verb) (increase(apiserver_request_total{job=\"apiserver\",verb=~\"LIST|GET|POST|PUT|PATCH|DELETE\",code=~\"5..\"}[1h]))\n", + "record": "code_verb:apiserver_request_total:increase1h" } ] }, @@ -485,8 +376,8 @@ data: "name": "k8s.rules", "rules": [ { - "expr": "sum by (cluster, namespace, pod, container) (\n rate(container_cpu_usage_seconds_total{job=\"kubernetes-cadvisor\", image!=\"\", container!=\"POD\"}[5m])\n) * on (cluster, namespace, pod) group_left(node) topk by (cluster, namespace, pod) (\n 1, max by(cluster, namespace, pod, node) (kube_pod_info{node!=\"\"})\n)\n", - "record": "node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate" + "expr": "sum by (cluster, namespace, pod, container) (\n irate(container_cpu_usage_seconds_total{job=\"kubernetes-cadvisor\", image!=\"\"}[5m])\n) * on (cluster, namespace, pod) group_left(node) topk by (cluster, namespace, pod) (\n 1, max by(cluster, namespace, pod, node) (kube_pod_info{node!=\"\"})\n)\n", + "record": "node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate" }, { "expr": "container_memory_working_set_bytes{job=\"kubernetes-cadvisor\", image!=\"\"}\n* on (namespace, pod) group_left(node) topk by(namespace, pod) (1,\n max by(namespace, pod, node) (kube_pod_info{node!=\"\"})\n)\n", @@ -505,12 +396,36 @@ data: "record": "node_namespace_pod_container:container_memory_swap" }, { - "expr": "sum by (namespace) (\n sum by (namespace, pod) (\n max by (namespace, pod, container) (\n kube_pod_container_resource_requests_memory_bytes{job=\"kube-state-metrics\"}\n ) * on(namespace, pod) group_left() max by (namespace, pod) (\n kube_pod_status_phase{phase=~\"Pending|Running\"} == 1\n )\n )\n)\n", - "record": "namespace:kube_pod_container_resource_requests_memory_bytes:sum" + "expr": "kube_pod_container_resource_requests{resource=\"memory\",job=\"kube-state-metrics\"} * on (namespace, pod, cluster)\ngroup_left() max by (namespace, pod, cluster) (\n (kube_pod_status_phase{phase=~\"Pending|Running\"} == 1)\n)\n", + "record": "cluster:namespace:pod_memory:active:kube_pod_container_resource_requests" }, { - "expr": "sum by (namespace) (\n sum by (namespace, pod) (\n max by (namespace, pod, container) (\n kube_pod_container_resource_requests_cpu_cores{job=\"kube-state-metrics\"}\n ) * on(namespace, pod) group_left() max by (namespace, pod) (\n kube_pod_status_phase{phase=~\"Pending|Running\"} == 1\n )\n )\n)\n", - "record": "namespace:kube_pod_container_resource_requests_cpu_cores:sum" + "expr": "sum by (namespace, cluster) (\n sum by (namespace, pod, cluster) (\n max by (namespace, pod, container, cluster) (\n kube_pod_container_resource_requests{resource=\"memory\",job=\"kube-state-metrics\"}\n ) * on(namespace, pod, cluster) group_left() max by (namespace, pod, cluster) (\n kube_pod_status_phase{phase=~\"Pending|Running\"} == 1\n )\n )\n)\n", + "record": "namespace_memory:kube_pod_container_resource_requests:sum" + }, + { + "expr": "kube_pod_container_resource_requests{resource=\"cpu\",job=\"kube-state-metrics\"} * on (namespace, pod, cluster)\ngroup_left() max by (namespace, pod, cluster) (\n (kube_pod_status_phase{phase=~\"Pending|Running\"} == 1)\n)\n", + "record": "cluster:namespace:pod_cpu:active:kube_pod_container_resource_requests" + }, + { + "expr": "sum by (namespace, cluster) (\n sum by (namespace, pod, cluster) (\n max by (namespace, pod, container, cluster) (\n kube_pod_container_resource_requests{resource=\"cpu\",job=\"kube-state-metrics\"}\n ) * on(namespace, pod, cluster) group_left() max by (namespace, pod, cluster) (\n kube_pod_status_phase{phase=~\"Pending|Running\"} == 1\n )\n )\n)\n", + "record": "namespace_cpu:kube_pod_container_resource_requests:sum" + }, + { + "expr": "kube_pod_container_resource_limits{resource=\"memory\",job=\"kube-state-metrics\"} * on (namespace, pod, cluster)\ngroup_left() max by (namespace, pod, cluster) (\n (kube_pod_status_phase{phase=~\"Pending|Running\"} == 1)\n)\n", + "record": "cluster:namespace:pod_memory:active:kube_pod_container_resource_limits" + }, + { + "expr": "sum by (namespace, cluster) (\n sum by (namespace, pod, cluster) (\n max by (namespace, pod, container, cluster) (\n kube_pod_container_resource_limits{resource=\"memory\",job=\"kube-state-metrics\"}\n ) * on(namespace, pod, cluster) group_left() max by (namespace, pod, cluster) (\n kube_pod_status_phase{phase=~\"Pending|Running\"} == 1\n )\n )\n)\n", + "record": "namespace_memory:kube_pod_container_resource_limits:sum" + }, + { + "expr": "kube_pod_container_resource_limits{resource=\"cpu\",job=\"kube-state-metrics\"} * on (namespace, pod, cluster)\ngroup_left() max by (namespace, pod, cluster) (\n (kube_pod_status_phase{phase=~\"Pending|Running\"} == 1)\n )\n", + "record": "cluster:namespace:pod_cpu:active:kube_pod_container_resource_limits" + }, + { + "expr": "sum by (namespace, cluster) (\n sum by (namespace, pod, cluster) (\n max by (namespace, pod, container, cluster) (\n kube_pod_container_resource_limits{resource=\"cpu\",job=\"kube-state-metrics\"}\n ) * on(namespace, pod, cluster) group_left() max by (namespace, pod, cluster) (\n kube_pod_status_phase{phase=~\"Pending|Running\"} == 1\n )\n )\n)\n", + "record": "namespace_cpu:kube_pod_container_resource_limits:sum" }, { "expr": "max by (cluster, namespace, workload, pod) (\n label_replace(\n label_replace(\n kube_pod_owner{job=\"kube-state-metrics\", owner_kind=\"ReplicaSet\"},\n \"replicaset\", \"$1\", \"owner_name\", \"(.*)\"\n ) * on(replicaset, namespace) group_left(owner_name) topk by(replicaset, namespace) (\n 1, max by (replicaset, namespace, owner_name) (\n kube_replicaset_owner{job=\"kube-state-metrics\"}\n )\n ),\n \"workload\", \"$1\", \"owner_name\", \"(.*)\"\n )\n)\n", @@ -532,6 +447,13 @@ data: "workload_type": "statefulset" }, "record": "namespace_workload_pod:kube_pod_owner:relabel" + }, + { + "expr": "max by (cluster, namespace, workload, pod) (\n label_replace(\n kube_pod_owner{job=\"kube-state-metrics\", owner_kind=\"Job\"},\n \"workload\", \"$1\", \"owner_name\", \"(.*)\"\n )\n)\n", + "labels": { + "workload_type": "job" + }, + "record": "namespace_workload_pod:kube_pod_owner:relabel" } ] }, @@ -611,12 +533,16 @@ data: "record": "node_namespace_pod:kube_pod_info:" }, { - "expr": "count by (cluster, node) (sum by (node, cpu) (\n node_cpu_seconds_total{job=\"node-exporter\"}\n* on (namespace, pod) group_left(node)\n node_namespace_pod:kube_pod_info:\n))\n", + "expr": "count by (cluster, node) (sum by (node, cpu) (\n node_cpu_seconds_total{job=\"node-exporter\"}\n* on (namespace, pod) group_left(node)\n topk by(namespace, pod) (1, node_namespace_pod:kube_pod_info:)\n))\n", "record": "node:node_num_cpu:sum" }, { "expr": "sum(\n node_memory_MemAvailable_bytes{job=\"node-exporter\"} or\n (\n node_memory_Buffers_bytes{job=\"node-exporter\"} +\n node_memory_Cached_bytes{job=\"node-exporter\"} +\n node_memory_MemFree_bytes{job=\"node-exporter\"} +\n node_memory_Slab_bytes{job=\"node-exporter\"}\n )\n) by (cluster)\n", "record": ":node_memory_MemAvailable_bytes:sum" + }, + { + "expr": "sum(rate(node_cpu_seconds_total{job=\"node-exporter\",mode!=\"idle\",mode!=\"iowait\",mode!=\"steal\"}[5m])) /\ncount(sum(node_cpu_seconds_total{job=\"node-exporter\"}) by (cluster, instance, cpu))\n", + "record": "cluster:node_cpu:ratio_rate5m" } ] }, @@ -624,21 +550,21 @@ data: "name": "kubelet.rules", "rules": [ { - "expr": "histogram_quantile(0.99, sum(rate(kubelet_pleg_relist_duration_seconds_bucket[5m])) by (instance, le) * on(instance) group_left(node) kubelet_node_name{job=\"kubelet\"})\n", + "expr": "histogram_quantile(0.99, sum(rate(kubelet_pleg_relist_duration_seconds_bucket[5m])) by (cluster, instance, le) * on(cluster, instance) group_left(node) kubelet_node_name{job=\"kubelet\"})\n", "labels": { "quantile": "0.99" }, "record": "node_quantile:kubelet_pleg_relist_duration_seconds:histogram_quantile" }, { - "expr": "histogram_quantile(0.9, sum(rate(kubelet_pleg_relist_duration_seconds_bucket[5m])) by (instance, le) * on(instance) group_left(node) kubelet_node_name{job=\"kubelet\"})\n", + "expr": "histogram_quantile(0.9, sum(rate(kubelet_pleg_relist_duration_seconds_bucket[5m])) by (cluster, instance, le) * on(cluster, instance) group_left(node) kubelet_node_name{job=\"kubelet\"})\n", "labels": { "quantile": "0.9" }, "record": "node_quantile:kubelet_pleg_relist_duration_seconds:histogram_quantile" }, { - "expr": "histogram_quantile(0.5, sum(rate(kubelet_pleg_relist_duration_seconds_bucket[5m])) by (instance, le) * on(instance) group_left(node) kubelet_node_name{job=\"kubelet\"})\n", + "expr": "histogram_quantile(0.5, sum(rate(kubelet_pleg_relist_duration_seconds_bucket[5m])) by (cluster, instance, le) * on(cluster, instance) group_left(node) kubelet_node_name{job=\"kubelet\"})\n", "labels": { "quantile": "0.5" }, @@ -652,11 +578,11 @@ data: { "alert": "KubePodCrashLooping", "annotations": { - "description": "Pod {{ $labels.namespace }}/{{ $labels.pod }} ({{ $labels.container }}) is restarting {{ printf \"%.2f\" $value }} times / 5 minutes.", + "description": "Pod {{ $labels.namespace }}/{{ $labels.pod }} ({{ $labels.container }}) is in waiting state (reason: \"CrashLoopBackOff\").", "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubepodcrashlooping", "summary": "Pod is crash looping." }, - "expr": "rate(kube_pod_container_status_restarts_total{job=\"kube-state-metrics\"}[5m]) * 60 * 5 > 0\n", + "expr": "max_over_time(kube_pod_container_status_waiting_reason{reason=\"CrashLoopBackOff\", job=\"kube-state-metrics\"}[5m]) >= 1\n", "for": "15m", "labels": { "severity": "warning" @@ -695,7 +621,7 @@ data: "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubedeploymentreplicasmismatch", "summary": "Deployment has not matched the expected number of replicas." }, - "expr": "(\n kube_deployment_spec_replicas{job=\"kube-state-metrics\"}\n !=\n kube_deployment_status_replicas_available{job=\"kube-state-metrics\"}\n) and (\n changes(kube_deployment_status_replicas_updated{job=\"kube-state-metrics\"}[5m])\n ==\n 0\n)\n", + "expr": "(\n kube_deployment_spec_replicas{job=\"kube-state-metrics\"}\n >\n kube_deployment_status_replicas_available{job=\"kube-state-metrics\"}\n) and (\n changes(kube_deployment_status_replicas_updated{job=\"kube-state-metrics\"}[10m])\n ==\n 0\n)\n", "for": "15m", "labels": { "severity": "warning" @@ -708,7 +634,7 @@ data: "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubestatefulsetreplicasmismatch", "summary": "Deployment has not matched the expected number of replicas." }, - "expr": "(\n kube_statefulset_status_replicas_ready{job=\"kube-state-metrics\"}\n !=\n kube_statefulset_status_replicas{job=\"kube-state-metrics\"}\n) and (\n changes(kube_statefulset_status_replicas_updated{job=\"kube-state-metrics\"}[5m])\n ==\n 0\n)\n", + "expr": "(\n kube_statefulset_status_replicas_ready{job=\"kube-state-metrics\"}\n !=\n kube_statefulset_status_replicas{job=\"kube-state-metrics\"}\n) and (\n changes(kube_statefulset_status_replicas_updated{job=\"kube-state-metrics\"}[10m])\n ==\n 0\n)\n", "for": "15m", "labels": { "severity": "warning" @@ -747,7 +673,7 @@ data: "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubedaemonsetrolloutstuck", "summary": "DaemonSet rollout is stuck." }, - "expr": "(\n (\n kube_daemonset_status_current_number_scheduled{job=\"kube-state-metrics\"}\n !=\n kube_daemonset_status_desired_number_scheduled{job=\"kube-state-metrics\"}\n ) or (\n kube_daemonset_status_number_misscheduled{job=\"kube-state-metrics\"}\n !=\n 0\n ) or (\n kube_daemonset_updated_number_scheduled{job=\"kube-state-metrics\"}\n !=\n kube_daemonset_status_desired_number_scheduled{job=\"kube-state-metrics\"}\n ) or (\n kube_daemonset_status_number_available{job=\"kube-state-metrics\"}\n !=\n kube_daemonset_status_desired_number_scheduled{job=\"kube-state-metrics\"}\n )\n) and (\n changes(kube_daemonset_updated_number_scheduled{job=\"kube-state-metrics\"}[5m])\n ==\n 0\n)\n", + "expr": "(\n (\n kube_daemonset_status_current_number_scheduled{job=\"kube-state-metrics\"}\n !=\n kube_daemonset_status_desired_number_scheduled{job=\"kube-state-metrics\"}\n ) or (\n kube_daemonset_status_number_misscheduled{job=\"kube-state-metrics\"}\n !=\n 0\n ) or (\n kube_daemonset_status_updated_number_scheduled{job=\"kube-state-metrics\"}\n !=\n kube_daemonset_status_desired_number_scheduled{job=\"kube-state-metrics\"}\n ) or (\n kube_daemonset_status_number_available{job=\"kube-state-metrics\"}\n !=\n kube_daemonset_status_desired_number_scheduled{job=\"kube-state-metrics\"}\n )\n) and (\n changes(kube_daemonset_status_updated_number_scheduled{job=\"kube-state-metrics\"}[5m])\n ==\n 0\n)\n", "for": "15m", "labels": { "severity": "warning" @@ -756,7 +682,7 @@ data: { "alert": "KubeContainerWaiting", "annotations": { - "description": "Pod {{ $labels.namespace }}/{{ $labels.pod }} container {{ $labels.container}} has been in waiting state for longer than 1 hour.", + "description": "pod/{{ $labels.pod }} in namespace {{ $labels.namespace }} on container {{ $labels.container}} has been in waiting state for longer than 1 hour.", "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubecontainerwaiting", "summary": "Pod container waiting longer than 1 hour" }, @@ -821,11 +747,11 @@ data: { "alert": "KubeHpaReplicasMismatch", "annotations": { - "description": "HPA {{ $labels.namespace }}/{{ $labels.hpa }} has not matched the desired number of replicas for longer than 15 minutes.", + "description": "HPA {{ $labels.namespace }}/{{ $labels.horizontalpodautoscaler }} has not matched the desired number of replicas for longer than 15 minutes.", "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubehpareplicasmismatch", "summary": "HPA has not matched descired number of replicas." }, - "expr": "(kube_hpa_status_desired_replicas{job=\"kube-state-metrics\"}\n !=\nkube_hpa_status_current_replicas{job=\"kube-state-metrics\"})\n and\n(kube_hpa_status_current_replicas{job=\"kube-state-metrics\"}\n >\nkube_hpa_spec_min_replicas{job=\"kube-state-metrics\"})\n and\n(kube_hpa_status_current_replicas{job=\"kube-state-metrics\"}\n <\nkube_hpa_spec_max_replicas{job=\"kube-state-metrics\"})\n and\nchanges(kube_hpa_status_current_replicas[15m]) == 0\n", + "expr": "(kube_horizontalpodautoscaler_status_desired_replicas{job=\"kube-state-metrics\"}\n !=\nkube_horizontalpodautoscaler_status_current_replicas{job=\"kube-state-metrics\"})\n and\n(kube_horizontalpodautoscaler_status_current_replicas{job=\"kube-state-metrics\"}\n >\nkube_horizontalpodautoscaler_spec_min_replicas{job=\"kube-state-metrics\"})\n and\n(kube_horizontalpodautoscaler_status_current_replicas{job=\"kube-state-metrics\"}\n <\nkube_horizontalpodautoscaler_spec_max_replicas{job=\"kube-state-metrics\"})\n and\nchanges(kube_horizontalpodautoscaler_status_current_replicas{job=\"kube-state-metrics\"}[15m]) == 0\n", "for": "15m", "labels": { "severity": "warning" @@ -834,11 +760,11 @@ data: { "alert": "KubeHpaMaxedOut", "annotations": { - "description": "HPA {{ $labels.namespace }}/{{ $labels.hpa }} has been running at max replicas for longer than 15 minutes.", + "description": "HPA {{ $labels.namespace }}/{{ $labels.horizontalpodautoscaler }} has been running at max replicas for longer than 15 minutes.", "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubehpamaxedout", "summary": "HPA is running at max replicas" }, - "expr": "kube_hpa_status_current_replicas{job=\"kube-state-metrics\"}\n ==\nkube_hpa_spec_max_replicas{job=\"kube-state-metrics\"}\n", + "expr": "kube_horizontalpodautoscaler_status_current_replicas{job=\"kube-state-metrics\"}\n ==\nkube_horizontalpodautoscaler_spec_max_replicas{job=\"kube-state-metrics\"}\n", "for": "15m", "labels": { "severity": "warning" @@ -852,12 +778,12 @@ data: { "alert": "KubeCPUOvercommit", "annotations": { - "description": "Cluster has overcommitted CPU resource requests for Pods and cannot tolerate node failure.", + "description": "Cluster has overcommitted CPU resource requests for Pods by {{ $value }} CPU shares and cannot tolerate node failure.", "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubecpuovercommit", "summary": "Cluster has overcommitted CPU resource requests." }, - "expr": "sum(namespace:kube_pod_container_resource_requests_cpu_cores:sum{})\n /\nsum(kube_node_status_allocatable_cpu_cores)\n >\n(count(kube_node_status_allocatable_cpu_cores)-1) / count(kube_node_status_allocatable_cpu_cores)\n", - "for": "5m", + "expr": "sum(namespace_cpu:kube_pod_container_resource_requests:sum{}) - (sum(kube_node_status_allocatable{resource=\"cpu\"}) - max(kube_node_status_allocatable{resource=\"cpu\"})) > 0\nand\n(sum(kube_node_status_allocatable{resource=\"cpu\"}) - max(kube_node_status_allocatable{resource=\"cpu\"})) > 0\n", + "for": "10m", "labels": { "severity": "warning" } @@ -865,12 +791,12 @@ data: { "alert": "KubeMemoryOvercommit", "annotations": { - "description": "Cluster has overcommitted memory resource requests for Pods and cannot tolerate node failure.", + "description": "Cluster has overcommitted memory resource requests for Pods by {{ $value | humanize }} bytes and cannot tolerate node failure.", "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubememoryovercommit", "summary": "Cluster has overcommitted memory resource requests." }, - "expr": "sum(namespace:kube_pod_container_resource_requests_memory_bytes:sum{})\n /\nsum(kube_node_status_allocatable_memory_bytes)\n >\n(count(kube_node_status_allocatable_memory_bytes)-1)\n /\ncount(kube_node_status_allocatable_memory_bytes)\n", - "for": "5m", + "expr": "sum(namespace_memory:kube_pod_container_resource_requests:sum{}) - (sum(kube_node_status_allocatable{resource=\"memory\"}) - max(kube_node_status_allocatable{resource=\"memory\"})) > 0\nand\n(sum(kube_node_status_allocatable{resource=\"memory\"}) - max(kube_node_status_allocatable{resource=\"memory\"})) > 0\n", + "for": "10m", "labels": { "severity": "warning" } @@ -882,7 +808,7 @@ data: "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubecpuquotaovercommit", "summary": "Cluster has overcommitted CPU resource requests." }, - "expr": "sum(kube_resourcequota{job=\"kube-state-metrics\", type=\"hard\", resource=\"cpu\"})\n /\nsum(kube_node_status_allocatable_cpu_cores)\n > 1.5\n", + "expr": "sum(min without(resource) (kube_resourcequota{job=\"kube-state-metrics\", type=\"hard\", resource=~\"(cpu|requests.cpu)\"}))\n /\nsum(kube_node_status_allocatable{resource=\"cpu\", job=\"kube-state-metrics\"})\n > 1.5\n", "for": "5m", "labels": { "severity": "warning" @@ -895,7 +821,7 @@ data: "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubememoryquotaovercommit", "summary": "Cluster has overcommitted memory resource requests." }, - "expr": "sum(kube_resourcequota{job=\"kube-state-metrics\", type=\"hard\", resource=\"memory\"})\n /\nsum(kube_node_status_allocatable_memory_bytes{job=\"kube-state-metrics\"})\n > 1.5\n", + "expr": "sum(min without(resource) (kube_resourcequota{job=\"kube-state-metrics\", type=\"hard\", resource=~\"(memory|requests.memory)\"}))\n /\nsum(kube_node_status_allocatable{resource=\"memory\", job=\"kube-state-metrics\"})\n > 1.5\n", "for": "5m", "labels": { "severity": "warning" @@ -965,7 +891,7 @@ data: "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubepersistentvolumefillingup", "summary": "PersistentVolume is filling up." }, - "expr": "kubelet_volume_stats_available_bytes{job=\"kubelet\"}\n /\nkubelet_volume_stats_capacity_bytes{job=\"kubelet\"}\n < 0.03\n", + "expr": "(\n kubelet_volume_stats_available_bytes{job=\"kubelet\"}\n /\n kubelet_volume_stats_capacity_bytes{job=\"kubelet\"}\n) < 0.03\nand\nkubelet_volume_stats_used_bytes{job=\"kubelet\"} > 0\nunless on(namespace, persistentvolumeclaim)\nkube_persistentvolumeclaim_access_mode{ access_mode=\"ReadOnlyMany\"} == 1\nunless on(namespace, persistentvolumeclaim)\nkube_persistentvolumeclaim_labels{label_excluded_from_alerts=\"true\"} == 1\n", "for": "1m", "labels": { "severity": "critical" @@ -978,7 +904,7 @@ data: "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubepersistentvolumefillingup", "summary": "PersistentVolume is filling up." }, - "expr": "(\n kubelet_volume_stats_available_bytes{job=\"kubelet\"}\n /\n kubelet_volume_stats_capacity_bytes{job=\"kubelet\"}\n) < 0.15\nand\npredict_linear(kubelet_volume_stats_available_bytes{job=\"kubelet\"}[6h], 4 * 24 * 3600) < 0\n", + "expr": "(\n kubelet_volume_stats_available_bytes{job=\"kubelet\"}\n /\n kubelet_volume_stats_capacity_bytes{job=\"kubelet\"}\n) < 0.15\nand\nkubelet_volume_stats_used_bytes{job=\"kubelet\"} > 0\nand\npredict_linear(kubelet_volume_stats_available_bytes{job=\"kubelet\"}[6h], 4 * 24 * 3600) < 0\nunless on(namespace, persistentvolumeclaim)\nkube_persistentvolumeclaim_access_mode{ access_mode=\"ReadOnlyMany\"} == 1\nunless on(namespace, persistentvolumeclaim)\nkube_persistentvolumeclaim_labels{label_excluded_from_alerts=\"true\"} == 1\n", "for": "1h", "labels": { "severity": "warning" @@ -1009,7 +935,7 @@ data: "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeversionmismatch", "summary": "Different semantic versions of Kubernetes components running." }, - "expr": "count(count by (gitVersion) (label_replace(kubernetes_build_info{job!~\"kube-dns|coredns\"},\"gitVersion\",\"$1\",\"gitVersion\",\"(v[0-9]*.[0-9]*).*\"))) > 1\n", + "expr": "count(count by (git_version) (label_replace(kubernetes_build_info{job!~\"kube-dns|coredns\"},\"git_version\",\"$1\",\"git_version\",\"(v[0-9]*.[0-9]*).*\"))) > 1\n", "for": "15m", "labels": { "severity": "warning" @@ -1022,7 +948,7 @@ data: "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeclienterrors", "summary": "Kubernetes API server client is experiencing errors." }, - "expr": "(sum(rate(rest_client_requests_total{code=~\"5..\"}[5m])) by (instance, job)\n /\nsum(rate(rest_client_requests_total[5m])) by (instance, job))\n> 0.01\n", + "expr": "(sum(rate(rest_client_requests_total{code=~\"5..\"}[5m])) by (instance, job, namespace)\n /\nsum(rate(rest_client_requests_total[5m])) by (instance, job, namespace))\n> 0.01\n", "for": "15m", "labels": { "severity": "warning" @@ -1101,7 +1027,7 @@ data: { "alert": "KubeClientCertificateExpiration", "annotations": { - "description": "A client certificate used to authenticate to the apiserver is expiring in less than 1.0 hours.", + "description": "A client certificate used to authenticate to kubernetes apiserver is expiring in less than 1.0 hours.", "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeclientcertificateexpiration", "summary": "Client certificate is about to expire." }, @@ -1113,7 +1039,7 @@ data: { "alert": "KubeClientCertificateExpiration", "annotations": { - "description": "A client certificate used to authenticate to the apiserver is expiring in less than 0.1 hours.", + "description": "A client certificate used to authenticate to kubernetes apiserver is expiring in less than 0.1 hours.", "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeclientcertificateexpiration", "summary": "Client certificate is about to expire." }, @@ -1123,23 +1049,23 @@ data: } }, { - "alert": "AggregatedAPIErrors", + "alert": "KubeAggregatedAPIErrors", "annotations": { - "description": "An aggregated API {{ $labels.name }}/{{ $labels.namespace }} has reported errors. It has appeared unavailable {{ $value | humanize }} times averaged over the past 10m.", - "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-aggregatedapierrors", - "summary": "An aggregated API has reported errors." + "description": "Kubernetes aggregated API {{ $labels.name }}/{{ $labels.namespace }} has reported errors. It has appeared unavailable {{ $value | humanize }} times averaged over the past 10m.", + "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeaggregatedapierrors", + "summary": "Kubernetes aggregated API has reported errors." }, - "expr": "sum by(name, namespace)(increase(aggregator_unavailable_apiservice_count[10m])) > 4\n", + "expr": "sum by(name, namespace)(increase(aggregator_unavailable_apiservice_total[10m])) > 4\n", "labels": { "severity": "warning" } }, { - "alert": "AggregatedAPIDown", + "alert": "KubeAggregatedAPIDown", "annotations": { - "description": "An aggregated API {{ $labels.name }}/{{ $labels.namespace }} has been only {{ $value | humanize }}% available over the last 10m.", - "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-aggregatedapidown", - "summary": "An aggregated API is down." + "description": "Kubernetes aggregated API {{ $labels.name }}/{{ $labels.namespace }} has been only {{ $value | humanize }}% available over the last 10m.", + "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeaggregatedapidown", + "summary": "Kubernetes aggregated API is down." }, "expr": "(1 - max by(name, namespace)(avg_over_time(aggregator_unavailable_apiservice[10m]))) * 100 < 85\n", "for": "5m", @@ -1159,6 +1085,19 @@ data: "labels": { "severity": "critical" } + }, + { + "alert": "KubeAPITerminatedRequests", + "annotations": { + "description": "The kubernetes apiserver has terminated {{ $value | humanizePercentage }} of its incoming requests.", + "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeapiterminatedrequests", + "summary": "The kubernetes apiserver has terminated {{ $value | humanizePercentage }} of its incoming requests." + }, + "expr": "sum(rate(apiserver_request_terminations_total{job=\"apiserver\"}[10m])) / ( sum(rate(apiserver_request_total{job=\"apiserver\"}[10m])) + sum(rate(apiserver_request_terminations_total{job=\"apiserver\"}[10m])) ) > 0.20\n", + "for": "5m", + "labels": { + "severity": "warning" + } } ] }, @@ -1198,10 +1137,10 @@ data: "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubelettoomanypods", "summary": "Kubelet is running at capacity." }, - "expr": "count by(node) (\n (kube_pod_status_phase{job=\"kube-state-metrics\",phase=\"Running\"} == 1) * on(instance,pod,namespace,cluster) group_left(node) topk by(instance,pod,namespace,cluster) (1, kube_pod_info{job=\"kube-state-metrics\"})\n)\n/\nmax by(node) (\n kube_node_status_capacity_pods{job=\"kube-state-metrics\"} != 1\n) > 0.95\n", + "expr": "count by(node) (\n (kube_pod_status_phase{job=\"kube-state-metrics\",phase=\"Running\"} == 1) * on(instance,pod,namespace,cluster) group_left(node) topk by(instance,pod,namespace,cluster) (1, kube_pod_info{job=\"kube-state-metrics\"})\n)\n/\nmax by(node) (\n kube_node_status_capacity{job=\"kube-state-metrics\",resource=\"pods\"} != 1\n) > 0.95\n", "for": "15m", "labels": { - "severity": "warning" + "severity": "info" } }, { @@ -1367,6 +1306,24 @@ data: } } ] + }, + { + "name": "kubernetes-system-kube-proxy", + "rules": [ + { + "alert": "KubeProxyDown", + "annotations": { + "description": "KubeProxy has disappeared from Prometheus target discovery.", + "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeproxydown", + "summary": "Target disappeared from Prometheus target discovery." + }, + "expr": "absent(up{job=\"kube-proxy\"} == 1)\n", + "for": "15m", + "labels": { + "severity": "critical" + } + } + ] } ] } @@ -1377,48 +1334,48 @@ data: "name": "node-exporter.rules", "rules": [ { - "expr": "count without (cpu) (\n count without (mode) (\n node_cpu_seconds_total{job=\"node-exporter\"}\n )\n)\n", + "expr": "count without (cpu, mode) (\n node_cpu_seconds_total{job=\"node-exporter\",mode=\"idle\"}\n)\n", "record": "instance:node_num_cpu:sum" }, { - "expr": "1 - avg without (cpu, mode) (\n rate(node_cpu_seconds_total{job=\"node-exporter\", mode=\"idle\"}[1m])\n)\n", - "record": "instance:node_cpu_utilisation:rate1m" + "expr": "1 - avg without (cpu) (\n sum without (mode) (rate(node_cpu_seconds_total{job=\"node-exporter\", mode=~\"idle|iowait|steal\"}[5m]))\n)\n", + "record": "instance:node_cpu_utilisation:rate5m" }, { "expr": "(\n node_load1{job=\"node-exporter\"}\n/\n instance:node_num_cpu:sum{job=\"node-exporter\"}\n)\n", "record": "instance:node_load1_per_cpu:ratio" }, { - "expr": "1 - (\n node_memory_MemAvailable_bytes{job=\"node-exporter\"}\n/\n node_memory_MemTotal_bytes{job=\"node-exporter\"}\n)\n", + "expr": "1 - (\n (\n node_memory_MemAvailable_bytes{job=\"node-exporter\"}\n or\n (\n node_memory_Buffers_bytes{job=\"node-exporter\"}\n +\n node_memory_Cached_bytes{job=\"node-exporter\"}\n +\n node_memory_MemFree_bytes{job=\"node-exporter\"}\n +\n node_memory_Slab_bytes{job=\"node-exporter\"}\n )\n )\n/\n node_memory_MemTotal_bytes{job=\"node-exporter\"}\n)\n", "record": "instance:node_memory_utilisation:ratio" }, { - "expr": "rate(node_vmstat_pgmajfault{job=\"node-exporter\"}[1m])\n", - "record": "instance:node_vmstat_pgmajfault:rate1m" + "expr": "rate(node_vmstat_pgmajfault{job=\"node-exporter\"}[5m])\n", + "record": "instance:node_vmstat_pgmajfault:rate5m" }, { - "expr": "rate(node_disk_io_time_seconds_total{job=\"node-exporter\", device!~\"dm.*\"}[1m])\n", - "record": "instance_device:node_disk_io_time_seconds:rate1m" + "expr": "rate(node_disk_io_time_seconds_total{job=\"node-exporter\", device!~\"dm.*\"}[5m])\n", + "record": "instance_device:node_disk_io_time_seconds:rate5m" }, { - "expr": "rate(node_disk_io_time_weighted_seconds_total{job=\"node-exporter\", device!~\"dm.*\"}[1m])\n", - "record": "instance_device:node_disk_io_time_weighted_seconds:rate1m" + "expr": "rate(node_disk_io_time_weighted_seconds_total{job=\"node-exporter\", device!~\"dm.*\"}[5m])\n", + "record": "instance_device:node_disk_io_time_weighted_seconds:rate5m" }, { - "expr": "sum without (device) (\n rate(node_network_receive_bytes_total{job=\"node-exporter\", device!=\"lo\"}[1m])\n)\n", - "record": "instance:node_network_receive_bytes_excluding_lo:rate1m" + "expr": "sum without (device) (\n rate(node_network_receive_bytes_total{job=\"node-exporter\", device!=\"lo\"}[5m])\n)\n", + "record": "instance:node_network_receive_bytes_excluding_lo:rate5m" }, { - "expr": "sum without (device) (\n rate(node_network_transmit_bytes_total{job=\"node-exporter\", device!=\"lo\"}[1m])\n)\n", - "record": "instance:node_network_transmit_bytes_excluding_lo:rate1m" + "expr": "sum without (device) (\n rate(node_network_transmit_bytes_total{job=\"node-exporter\", device!=\"lo\"}[5m])\n)\n", + "record": "instance:node_network_transmit_bytes_excluding_lo:rate5m" }, { - "expr": "sum without (device) (\n rate(node_network_receive_drop_total{job=\"node-exporter\", device!=\"lo\"}[1m])\n)\n", - "record": "instance:node_network_receive_drop_excluding_lo:rate1m" + "expr": "sum without (device) (\n rate(node_network_receive_drop_total{job=\"node-exporter\", device!=\"lo\"}[5m])\n)\n", + "record": "instance:node_network_receive_drop_excluding_lo:rate5m" }, { - "expr": "sum without (device) (\n rate(node_network_transmit_drop_total{job=\"node-exporter\", device!=\"lo\"}[1m])\n)\n", - "record": "instance:node_network_transmit_drop_excluding_lo:rate1m" + "expr": "sum without (device) (\n rate(node_network_transmit_drop_total{job=\"node-exporter\", device!=\"lo\"}[5m])\n)\n", + "record": "instance:node_network_transmit_drop_excluding_lo:rate5m" } ] }, @@ -1456,7 +1413,7 @@ data: "summary": "Filesystem has less than 5% space left." }, "expr": "(\n node_filesystem_avail_bytes{job=\"node-exporter\",fstype!~\"tmpfs|nsfs|vfat\"} / node_filesystem_size_bytes{job=\"node-exporter\",fstype!~\"tmpfs|nsfs|vfat\"} * 100 < 5\nand\n node_filesystem_readonly{job=\"node-exporter\",fstype!~\"tmpfs|nsfs|vfat\"} == 0\n)\n", - "for": "1h", + "for": "30m", "labels": { "severity": "warning" } @@ -1468,7 +1425,7 @@ data: "summary": "Filesystem has less than 3% space left." }, "expr": "(\n node_filesystem_avail_bytes{job=\"node-exporter\",fstype!~\"tmpfs|nsfs|vfat\"} / node_filesystem_size_bytes{job=\"node-exporter\",fstype!~\"tmpfs|nsfs|vfat\"} * 100 < 3\nand\n node_filesystem_readonly{job=\"node-exporter\",fstype!~\"tmpfs|nsfs|vfat\"} == 0\n)\n", - "for": "1h", + "for": "30m", "labels": { "severity": "critical" } @@ -1570,7 +1527,7 @@ data: { "alert": "NodeClockSkewDetected", "annotations": { - "message": "Clock on {{ $labels.instance }} is out of sync by more than 300s. Ensure NTP is configured correctly on this host.", + "description": "Clock on {{ $labels.instance }} is out of sync by more than 300s. Ensure NTP is configured correctly on this host.", "summary": "Clock skew detected." }, "expr": "(\n node_timex_offset_seconds > 0.05\nand\n deriv(node_timex_offset_seconds[5m]) >= 0\n)\nor\n(\n node_timex_offset_seconds < -0.05\nand\n deriv(node_timex_offset_seconds[5m]) <= 0\n)\n", @@ -1582,7 +1539,7 @@ data: { "alert": "NodeClockNotSynchronising", "annotations": { - "message": "Clock on {{ $labels.instance }} is not synchronising. Ensure NTP is configured on this host.", + "description": "Clock on {{ $labels.instance }} is not synchronising. Ensure NTP is configured on this host.", "summary": "Clock not synchronising." }, "expr": "min_over_time(node_timex_sync_status[5m]) == 0\nand\nnode_timex_maxerror_seconds >= 16\n", @@ -1609,10 +1566,34 @@ data: "description": "At least one device in RAID array on {{ $labels.instance }} failed. Array '{{ $labels.device }}' needs attention and possibly a disk swap.", "summary": "Failed device in RAID array" }, - "expr": "node_md_disks{state=\"fail\"} > 0\n", + "expr": "node_md_disks{state=\"failed\"} > 0\n", "labels": { "severity": "warning" } + }, + { + "alert": "NodeFileDescriptorLimit", + "annotations": { + "description": "File descriptors limit at {{ $labels.instance }} is currently at {{ printf \"%.2f\" $value }}%.", + "summary": "Kernel is predicted to exhaust file descriptors limit soon." + }, + "expr": "(\n node_filefd_allocated{job=\"node-exporter\"} * 100 / node_filefd_maximum{job=\"node-exporter\"} > 70\n)\n", + "for": "15m", + "labels": { + "severity": "warning" + } + }, + { + "alert": "NodeFileDescriptorLimit", + "annotations": { + "description": "File descriptors limit at {{ $labels.instance }} is currently at {{ printf \"%.2f\" $value }}%.", + "summary": "Kernel is predicted to exhaust file descriptors limit soon." + }, + "expr": "(\n node_filefd_allocated{job=\"node-exporter\"} * 100 / node_filefd_maximum{job=\"node-exporter\"} > 90\n)\n", + "for": "15m", + "labels": { + "severity": "critical" + } } ] } @@ -1738,7 +1719,7 @@ data: "description": "Prometheus {{$labels.instance}} failed to send {{ printf \"%.1f\" $value }}% of the samples to {{ $labels.remote_name}}:{{ $labels.url }}", "summary": "Prometheus fails to send samples to remote storage." }, - "expr": "(\n rate(prometheus_remote_storage_failed_samples_total{job=\"prometheus\"}[5m])\n/\n (\n rate(prometheus_remote_storage_failed_samples_total{job=\"prometheus\"}[5m])\n +\n rate(prometheus_remote_storage_succeeded_samples_total{job=\"prometheus\"}[5m])\n )\n)\n* 100\n> 1\n", + "expr": "(\n (rate(prometheus_remote_storage_failed_samples_total{job=\"prometheus\"}[5m]) or rate(prometheus_remote_storage_samples_failed_total{job=\"prometheus\"}[5m]))\n/\n (\n (rate(prometheus_remote_storage_failed_samples_total{job=\"prometheus\"}[5m]) or rate(prometheus_remote_storage_samples_failed_total{job=\"prometheus\"}[5m]))\n +\n (rate(prometheus_remote_storage_succeeded_samples_total{job=\"prometheus\"}[5m]) or rate(prometheus_remote_storage_samples_total{job=\"prometheus\"}[5m]))\n )\n)\n* 100\n> 1\n", "for": "15m", "labels": { "severity": "critical" @@ -1804,6 +1785,30 @@ data: "severity": "warning" } }, + { + "alert": "PrometheusLabelLimitHit", + "annotations": { + "description": "Prometheus {{$labels.instance}} has dropped {{ printf \"%.0f\" $value }} targets because some samples exceeded the configured label_limit, label_name_length_limit or label_value_length_limit.", + "summary": "Prometheus has dropped targets because some scrape configs have exceeded the labels limit." + }, + "expr": "increase(prometheus_target_scrape_pool_exceeded_label_limits_total{job=\"prometheus\"}[5m]) > 0\n", + "for": "15m", + "labels": { + "severity": "warning" + } + }, + { + "alert": "PrometheusTargetSyncFailure", + "annotations": { + "description": "{{ printf \"%.0f\" $value }} targets in Prometheus {{$labels.instance}} have failed to sync because invalid configuration was supplied.", + "summary": "Prometheus has failed to sync targets." + }, + "expr": "increase(prometheus_target_sync_failed_total{job=\"prometheus\"}[30m]) > 0\n", + "for": "5m", + "labels": { + "severity": "critical" + } + }, { "alert": "PrometheusErrorSendingAlertsToAnyAlertmanager", "annotations": {