From 68d8717924a16d8ea6306b1ef9a401fdb0020cd6 Mon Sep 17 00:00:00 2001 From: Dalton Hubble Date: Sun, 21 Jul 2019 11:02:18 -0700 Subject: [PATCH] Refresh Prometheus rules/alerts and Grafana dashboards * Refresh rules, alerts, and dashboards from upstreams --- CHANGES.md | 1 + addons/grafana/dashboards-etcd.yaml | 2 +- addons/grafana/dashboards-k8s-nodes.yaml | 5005 +++++++++++++++++ addons/grafana/dashboards-k8s-resources.yaml | 2035 ++++++- addons/grafana/dashboards-k8s.yaml | 5221 +++++++++--------- addons/grafana/dashboards-prom.yaml | 2160 ++++++++ addons/grafana/deployment.yaml | 10 + addons/prometheus/rules.yaml | 417 +- 8 files changed, 12083 insertions(+), 2768 deletions(-) create mode 100644 addons/grafana/dashboards-k8s-nodes.yaml create mode 100644 addons/grafana/dashboards-prom.yaml diff --git a/CHANGES.md b/CHANGES.md index f0bef47f..99d35df1 100644 --- a/CHANGES.md +++ b/CHANGES.md @@ -25,6 +25,7 @@ Notable changes between versions. #### Addons * Update Prometheus from v2.10.0 to v2.11.0 + * Refresh rules, alerts, and dashboards from upstreams * Update kube-state-metrics from v1.6.0 to v1.7.1 * Update Grafana from v6.2.4 to v6.2.5 * Update nginx-ingress from v0.24.1 to v0.25.0 diff --git a/addons/grafana/dashboards-etcd.yaml b/addons/grafana/dashboards-etcd.yaml index 21f9880f..94bcaf06 100644 --- a/addons/grafana/dashboards-etcd.yaml +++ b/addons/grafana/dashboards-etcd.yaml @@ -19,7 +19,7 @@ data: "links": [ ], - "refresh": false, + "refresh": "10s", "rows": [ { "collapse": false, diff --git a/addons/grafana/dashboards-k8s-nodes.yaml b/addons/grafana/dashboards-k8s-nodes.yaml new file mode 100644 index 00000000..a0d7963c --- /dev/null +++ b/addons/grafana/dashboards-k8s-nodes.yaml @@ -0,0 +1,5005 @@ +apiVersion: v1 +kind: ConfigMap +metadata: + name: grafana-dashboards-k8s-nodes + namespace: monitoring +data: + kubelet.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": "$datasource", + "format": "none", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "gridPos": { + + }, + "id": 2, + "interval": null, + "links": [ + + ], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 2, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "sum(up{job=\"kubelet\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "", + "refId": "A" + } + ], + "thresholds": "", + "title": "Up", + "tooltip": { + "shared": false + }, + "type": "singlestat", + "valueFontSize": "80%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "min" + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": [ + "#299c46", + "rgba(237, 129, 40, 0.89)", + "#d44a3a" + ], + "datasource": "$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_pod_count{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_container_count{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{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{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{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, + "gridPos": { + + }, + "id": 8, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "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{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, + "gridPos": { + + }, + "id": 9, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "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{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, + "gridPos": { + + }, + "id": 10, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "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{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, + "gridPos": { + + }, + "id": 11, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "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{job=\"kubelet\",instance=~\"$instance\"}[5m])) by (instance)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} pod", + "refId": "A" + }, + { + "expr": "sum(rate(kubelet_pod_worker_duration_seconds_count{job=\"kubelet\",instance=~\"$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": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 12, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "total": false, + "values": "true" + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.99, sum(rate(kubelet_pod_start_duration_seconds_count{job=\"kubelet\",instance=~\"$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{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, + "gridPos": { + + }, + "id": 13, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "hideEmpty": "true", + "hideZero": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "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{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, + "gridPos": { + + }, + "id": 14, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "hideEmpty": "true", + "hideZero": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "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{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, + "gridPos": { + + }, + "id": 15, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "hideEmpty": "true", + "hideZero": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": true, + "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{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, + "gridPos": { + + }, + "id": 16, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "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{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, + "gridPos": { + + }, + "id": 17, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "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{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, + "gridPos": { + + }, + "id": 18, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "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{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, + "gridPos": { + + }, + "id": 19, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "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{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, + "gridPos": { + + }, + "id": 20, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "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{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, + "gridPos": { + + }, + "id": 21, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "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{job=\"kubelet\", instance=~\"$instance\",code=~\"2..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "2xx", + "refId": "A" + }, + { + "expr": "sum(rate(rest_client_requests_total{job=\"kubelet\", instance=~\"$instance\",code=~\"3..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "3xx", + "refId": "B" + }, + { + "expr": "sum(rate(rest_client_requests_total{job=\"kubelet\", instance=~\"$instance\",code=~\"4..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "4xx", + "refId": "C" + }, + { + "expr": "sum(rate(rest_client_requests_total{job=\"kubelet\", instance=~\"$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": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 22, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "total": false, + "values": "true" + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_latency_seconds_bucket{job=\"kubelet\", instance=~\"$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": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 23, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 4, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "process_resident_memory_bytes{job=\"kubelet\",instance=~\"$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": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 24, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 4, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "rate(process_cpu_seconds_total{job=\"kubelet\",instance=~\"$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": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 25, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 4, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "go_goroutines{job=\"kubelet\",instance=~\"$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", + "tags": [ + "kubernetes-mixin" + ], + "templating": { + "list": [ + { + "current": { + "text": "Prometheus", + "value": "Prometheus" + }, + "hide": 0, + "label": null, + "name": "datasource", + "options": [ + + ], + "query": "prometheus", + "refresh": 1, + "regex": "", + "type": "datasource" + }, + { + "allValue": null, + "current": { + + }, + "datasource": "$datasource", + "hide": 0, + "includeAll": true, + "label": null, + "multi": false, + "name": "instance", + "options": [ + + ], + "query": "label_values(kubelet_runtime_operations{job=\"kubelet\"}, instance)", + "refresh": 2, + "regex": "", + "sort": 0, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + } + ] + }, + "time": { + "from": "now-1h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "5s", + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ], + "time_options": [ + "5m", + "15m", + "1h", + "6h", + "12h", + "24h", + "2d", + "7d", + "30d" + ] + }, + "timezone": "", + "title": "Kubernetes / Kubelet", + "uid": "3138fa155d5915769fbded898ac09fd9", + "version": 0 + } + nodes.json: |- + { + "__inputs": [ + + ], + "__requires": [ + + ], + "annotations": { + "list": [ + + ] + }, + "editable": false, + "gnetId": null, + "graphTooltip": 0, + "hideControls": false, + "id": null, + "links": [ + + ], + "refresh": "", + "rows": [ + { + "collapse": false, + "collapsed": false, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 2, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "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": 6, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "max(node_load1{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "load 1m", + "refId": "A" + }, + { + "expr": "max(node_load5{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "load 5m", + "refId": "B" + }, + { + "expr": "max(node_load15{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "load 15m", + "refId": "C" + }, + { + "expr": "count(node_cpu_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\", mode=\"user\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "logical cores", + "refId": "D" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "System load", + "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 + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 3, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum by (cpu) (irate(node_cpu_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", mode!=\"idle\", instance=\"$instance\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{cpu}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Usage Per Core", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "percentunit", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "percentunit", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Dashboard Row", + "titleSize": "h6", + "type": "row" + }, + { + "collapse": false, + "collapsed": false, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 4, + "legend": { + "alignAsTable": "true", + "avg": "true", + "current": "true", + "max": "false", + "min": "false", + "rightSide": "true", + "show": "true", + "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": 9, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "max (sum by (cpu) (irate(node_cpu_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", mode!=\"idle\", instance=\"$instance\"}[2m])) ) * 100\n", + "format": "time_series", + "intervalFactor": 10, + "legendFormat": "{{ cpu }}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "CPU Utilization", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "percent", + "label": null, + "logBase": 1, + "max": 100, + "min": 0, + "show": true + }, + { + "format": "percent", + "label": null, + "logBase": 1, + "max": 100, + "min": 0, + "show": true + } + ] + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": [ + "rgba(50, 172, 45, 0.97)", + "rgba(237, 129, 40, 0.89)", + "rgba(245, 54, 54, 0.9)" + ], + "datasource": "$datasource", + "format": "percent", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": true, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "gridPos": { + + }, + "id": 5, + "interval": null, + "links": [ + + ], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 3, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "avg(sum by (cpu) (irate(node_cpu_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", mode!=\"idle\", instance=\"$instance\"}[2m]))) * 100\n", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "", + "refId": "A" + } + ], + "thresholds": "80, 90", + "title": "CPU Usage", + "tooltip": { + "shared": false + }, + "type": "singlestat", + "valueFontSize": "80%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Dashboard Row", + "titleSize": "h6", + "type": "row" + }, + { + "collapse": false, + "collapsed": false, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 6, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "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": 9, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "max(\n node_memory_MemTotal_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_MemFree_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_Buffers_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_Cached_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n)\n", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "memory used", + "refId": "A" + }, + { + "expr": "max(node_memory_Buffers_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "memory buffers", + "refId": "B" + }, + { + "expr": "max(node_memory_Cached_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "memory cached", + "refId": "C" + }, + { + "expr": "max(node_memory_MemFree_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "memory free", + "refId": "D" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Memory Usage", + "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 + } + ] + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": [ + "rgba(50, 172, 45, 0.97)", + "rgba(237, 129, 40, 0.89)", + "rgba(245, 54, 54, 0.9)" + ], + "datasource": "$datasource", + "format": "percent", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": true, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "gridPos": { + + }, + "id": 7, + "interval": null, + "links": [ + + ], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 3, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "max(\n (\n (\n node_memory_MemTotal_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_MemFree_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_Buffers_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_Cached_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n )\n / node_memory_MemTotal_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n ) * 100)\n", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "", + "refId": "A" + } + ], + "thresholds": "80, 90", + "title": "Memory Usage", + "tooltip": { + "shared": false + }, + "type": "singlestat", + "valueFontSize": "80%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Dashboard Row", + "titleSize": "h6", + "type": "row" + }, + { + "collapse": false, + "collapsed": false, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 8, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + { + "alias": "read", + "yaxis": 1 + }, + { + "alias": "io time", + "yaxis": 2 + } + ], + "spaceLength": 10, + "span": 6, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "max(rate(node_disk_read_bytes_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}[2m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "read", + "refId": "A" + }, + { + "expr": "max(rate(node_disk_written_bytes_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}[2m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "written", + "refId": "B" + }, + { + "expr": "max(rate(node_disk_io_time_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}[2m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "io time", + "refId": "C" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Disk I/O", + "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": "ms", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 9, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "node:node_filesystem_usage:{cluster=\"$cluster\", instance=\"$instance\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{device}} disk used", + "refId": "A" + }, + { + "expr": "node:node_filesystem_usage:{cluster=\"$cluster\", instance=\"$instance\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{device}} disk free", + "refId": "B" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Disk Space Usage", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "percentunit", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "percentunit", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Dashboard Row", + "titleSize": "h6", + "type": "row" + }, + { + "collapse": false, + "collapsed": false, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 10, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "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": 6, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "rate(node_network_receive_bytes_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\", device!~\"lo\"}[5m])", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{device}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Network Received", + "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": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 11, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "rate(node_network_transmit_bytes_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\", device!~\"lo\"}[5m])", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{device}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Network Transmitted", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "bytes", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "bytes", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Dashboard Row", + "titleSize": "h6", + "type": "row" + }, + { + "collapse": false, + "collapsed": false, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 12, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "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": 9, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "max(\n node_filesystem_files{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_filesystem_files_free{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n)\n", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "inodes used", + "refId": "A" + }, + { + "expr": "max(node_filesystem_files_free{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "inodes free", + "refId": "B" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Inodes 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": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + }, + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": [ + "rgba(50, 172, 45, 0.97)", + "rgba(237, 129, 40, 0.89)", + "rgba(245, 54, 54, 0.9)" + ], + "datasource": "$datasource", + "format": "percent", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": true, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "gridPos": { + + }, + "id": 13, + "interval": null, + "links": [ + + ], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 3, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "max(\n (\n (\n node_filesystem_files{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_filesystem_files_free{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n )\n / node_filesystem_files{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n ) * 100)\n", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "", + "refId": "A" + } + ], + "thresholds": "80, 90", + "title": "Inodes Usage", + "tooltip": { + "shared": false + }, + "type": "singlestat", + "valueFontSize": "80%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "current" + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Dashboard Row", + "titleSize": "h6", + "type": "row" + } + ], + "schemaVersion": 14, + "style": "dark", + "tags": [ + "kubernetes-mixin" + ], + "templating": { + "list": [ + { + "current": { + "text": "Prometheus", + "value": "Prometheus" + }, + "hide": 0, + "label": null, + "name": "datasource", + "options": [ + + ], + "query": "prometheus", + "refresh": 1, + "regex": "", + "type": "datasource" + }, + { + "allValue": null, + "current": { + + }, + "datasource": "$datasource", + "hide": 2, + "includeAll": false, + "label": "cluster", + "multi": false, + "name": "cluster", + "options": [ + + ], + "query": "label_values(kube_pod_info, cluster)", + "refresh": 2, + "regex": "", + "sort": 0, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "current": { + + }, + "datasource": "$datasource", + "hide": 0, + "includeAll": false, + "label": null, + "multi": false, + "name": "instance", + "options": [ + + ], + "query": "label_values(node_boot_time_seconds{cluster=\"$cluster\", job=\"node-exporter\"}, instance)", + "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": "", + "title": "Kubernetes / Nodes", + "uid": "fa49a4706d07a042595b664c87fb33ea", + "version": 0 + } + proxy.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": "$datasource", + "format": "none", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "gridPos": { + + }, + "id": 2, + "interval": null, + "links": [ + + ], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 2, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "sum(up{job=\"kube-proxy\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "", + "refId": "A" + } + ], + "thresholds": "", + "title": "Up", + "tooltip": { + "shared": false + }, + "type": "singlestat", + "valueFontSize": "80%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "min" + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 3, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "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": 5, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum(rate(kubeproxy_sync_proxy_rules_duration_seconds_count{job=\"kube-proxy\", instance=~\"$instance\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "rate", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rules Sync Rate", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 4, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "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": 5, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.99,rate(kubeproxy_sync_proxy_rules_duration_seconds_bucket{job=\"kube-proxy\", instance=~\"$instance\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rule Sync Latency 99th Quantile", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Dashboard Row", + "titleSize": "h6", + "type": "row" + }, + { + "collapse": false, + "collapsed": false, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 5, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "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": 6, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum(rate(kubeproxy_network_programming_duration_seconds_count{job=\"kube-proxy\", instance=~\"$instance\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "rate", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Network Programming Rate", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 6, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "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(kubeproxy_network_programming_duration_seconds_bucket{job=\"kube-proxy\", instance=~\"$instance\"}[5m])) by (instance, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Network Programming Latency 99th Quantile", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Dashboard Row", + "titleSize": "h6", + "type": "row" + }, + { + "collapse": false, + "collapsed": false, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 7, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "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": "sum(rate(rest_client_requests_total{job=\"kube-proxy\", instance=~\"$instance\",code=~\"2..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "2xx", + "refId": "A" + }, + { + "expr": "sum(rate(rest_client_requests_total{job=\"kube-proxy\", instance=~\"$instance\",code=~\"3..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "3xx", + "refId": "B" + }, + { + "expr": "sum(rate(rest_client_requests_total{job=\"kube-proxy\", instance=~\"$instance\",code=~\"4..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "4xx", + "refId": "C" + }, + { + "expr": "sum(rate(rest_client_requests_total{job=\"kube-proxy\", instance=~\"$instance\",code=~\"5..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "5xx", + "refId": "D" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Kube API Request 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": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 8, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 8, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_latency_seconds_bucket{job=\"kube-proxy\",instance=~\"$instance\",verb=\"POST\"}[5m])) by (verb, url, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{verb}} {{url}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Post Request Latency 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": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 9, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "total": false, + "values": "true" + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_latency_seconds_bucket{job=\"kube-proxy\", instance=~\"$instance\", verb=\"GET\"}[5m])) by (verb, url, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{verb}} {{url}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Get Request Latency 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": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 10, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 4, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "process_resident_memory_bytes{job=\"kube-proxy\",instance=~\"$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": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 11, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 4, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "rate(process_cpu_seconds_total{job=\"kube-proxy\",instance=~\"$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": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 12, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 4, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "go_goroutines{job=\"kube-proxy\",instance=~\"$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", + "tags": [ + "kubernetes-mixin" + ], + "templating": { + "list": [ + { + "current": { + "text": "Prometheus", + "value": "Prometheus" + }, + "hide": 0, + "label": null, + "name": "datasource", + "options": [ + + ], + "query": "prometheus", + "refresh": 1, + "regex": "", + "type": "datasource" + }, + { + "allValue": null, + "current": { + + }, + "datasource": "$datasource", + "hide": 0, + "includeAll": true, + "label": null, + "multi": false, + "name": "instance", + "options": [ + + ], + "query": "label_values(kubeproxy_network_programming_duration_seconds_bucket{job=\"kube-proxy\"}, instance)", + "refresh": 2, + "regex": "", + "sort": 0, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + } + ] + }, + "time": { + "from": "now-1h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "5s", + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ], + "time_options": [ + "5m", + "15m", + "1h", + "6h", + "12h", + "24h", + "2d", + "7d", + "30d" + ] + }, + "timezone": "", + "title": "Kubernetes / Proxy", + "uid": "632e265de029684c40b21cb76bca4f94", + "version": 0 + } diff --git a/addons/grafana/dashboards-k8s-resources.yaml b/addons/grafana/dashboards-k8s-resources.yaml index f00af428..98a17712 100644 --- a/addons/grafana/dashboards-k8s-resources.yaml +++ b/addons/grafana/dashboards-k8s-resources.yaml @@ -4,6 +4,1907 @@ metadata: name: grafana-dashboards-k8s-resources namespace: monitoring data: + k8s-cluster-rsrc-use.json: |- + { + "annotations": { + "list": [ + + ] + }, + "editable": true, + "gnetId": null, + "graphTooltip": 0, + "hideControls": false, + "links": [ + + ], + "refresh": "10s", + "rows": [ + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 1, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "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": "node:cluster_cpu_utilisation:ratio{cluster=\"$cluster\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{node}}", + "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "CPU Utilisation", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "percentunit", + "label": null, + "logBase": 1, + "max": 1, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 2, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "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": "node:node_cpu_saturation_load1:{cluster=\"$cluster\"} / scalar(sum(min(kube_pod_info{cluster=\"$cluster\"}) by (node)))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{node}}", + "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "CPU Saturation (Load1)", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "percentunit", + "label": null, + "logBase": 1, + "max": 1, + "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": "CPU", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 3, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 0, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "node:cluster_memory_utilisation:ratio{cluster=\"$cluster\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{node}}", + "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Memory Utilisation", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "percentunit", + "label": null, + "logBase": 1, + "max": 1, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 4, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 0, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "node:node_memory_swap_io_bytes:sum_rate{cluster=\"$cluster\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{node}}", + "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Memory Saturation (Swap I/O)", + "tooltip": { + "shared": false, + "sort": 0, + "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": "Memory", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 5, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 0, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "node:node_disk_utilisation:avg_irate{cluster=\"$cluster\"} / scalar(:kube_pod_info_node_count:{cluster=\"$cluster\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{node}}", + "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Disk IO Utilisation", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "percentunit", + "label": null, + "logBase": 1, + "max": 1, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 6, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "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": "node:node_disk_saturation:avg_irate{cluster=\"$cluster\"} / scalar(:kube_pod_info_node_count:{cluster=\"$cluster\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{node}}", + "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Disk IO Saturation", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "percentunit", + "label": null, + "logBase": 1, + "max": 1, + "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": "Disk", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 7, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 0, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "node:node_net_utilisation:sum_irate{cluster=\"$cluster\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{node}}", + "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Net Utilisation (Transmitted)", + "tooltip": { + "shared": false, + "sort": 0, + "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 + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 8, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 0, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "node:node_net_saturation:sum_irate{cluster=\"$cluster\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{node}}", + "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Net Saturation (Dropped)", + "tooltip": { + "shared": false, + "sort": 0, + "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": "Network", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 9, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 0, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum(max(node_filesystem_size_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\", cluster=\"$cluster\"} - node_filesystem_avail_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\", cluster=\"$cluster\"}) by (device,pod,namespace)) by (pod,namespace)\n/ scalar(sum(max(node_filesystem_size_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\", cluster=\"$cluster\"}) by (device,pod,namespace)))\n* on (namespace, pod) group_left (node) node_namespace_pod:kube_pod_info:{cluster=\"$cluster\"}\n", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{node}}", + "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Disk Capacity", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "percentunit", + "label": null, + "logBase": 1, + "max": 1, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Storage", + "titleSize": "h6" + } + ], + "schemaVersion": 14, + "style": "dark", + "tags": [ + "kubernetes-mixin" + ], + "templating": { + "list": [ + { + "current": { + "text": "Prometheus", + "value": "Prometheus" + }, + "hide": 0, + "label": null, + "name": "datasource", + "options": [ + + ], + "query": "prometheus", + "refresh": 1, + "regex": "", + "type": "datasource" + }, + { + "allValue": null, + "current": { + "text": "prod", + "value": "prod" + }, + "datasource": "$datasource", + "hide": 2, + "includeAll": false, + "label": "cluster", + "multi": false, + "name": "cluster", + "options": [ + + ], + "query": "label_values(kube_node_info, cluster)", + "refresh": 1, + "regex": "", + "sort": 2, + "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": "", + "title": "Kubernetes / USE Method / Cluster", + "uid": "a6e7d1362e1ddbb79db21d5bb40d7137", + "version": 0 + } + k8s-node-rsrc-use.json: |- + { + "annotations": { + "list": [ + + ] + }, + "editable": true, + "gnetId": null, + "graphTooltip": 0, + "hideControls": false, + "links": [ + + ], + "refresh": "10s", + "rows": [ + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 1, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "node:node_cpu_utilisation:avg1m{cluster=\"$cluster\", node=\"$node\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "Utilisation", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "CPU Utilisation", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "percentunit", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 2, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "node:node_cpu_saturation_load1:{cluster=\"$cluster\", node=\"$node\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "Saturation", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "CPU Saturation (Load1)", + "tooltip": { + "shared": false, + "sort": 0, + "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": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "CPU", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 3, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "node:node_memory_utilisation:{cluster=\"$cluster\", node=\"$node\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "Memory", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Memory Utilisation", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "percentunit", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 4, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "node:node_memory_swap_io_bytes:sum_rate{cluster=\"$cluster\", node=\"$node\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "Swap IO", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Memory Saturation (Swap I/O)", + "tooltip": { + "shared": false, + "sort": 0, + "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": "Memory", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 5, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "node:node_disk_utilisation:avg_irate{cluster=\"$cluster\", node=\"$node\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "Utilisation", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Disk IO Utilisation", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "percentunit", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 6, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "node:node_disk_saturation:avg_irate{cluster=\"$cluster\", node=\"$node\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "Saturation", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Disk IO Saturation", + "tooltip": { + "shared": false, + "sort": 0, + "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": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Disk", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 7, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "node:node_net_utilisation:sum_irate{cluster=\"$cluster\", node=\"$node\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "Utilisation", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Net Utilisation (Transmitted)", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "Bps", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 8, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "node:node_net_saturation:sum_irate{cluster=\"$cluster\", node=\"$node\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "Saturation", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Net Saturation (Dropped)", + "tooltip": { + "shared": false, + "sort": 0, + "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": "Net", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 9, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "node:node_filesystem_usage:{cluster=\"$cluster\"}\n* on (namespace, pod) group_left (node) node_namespace_pod:kube_pod_info:{cluster=\"$cluster\", node=\"$node\"}\n", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{device}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Disk Utilisation", + "tooltip": { + "shared": false, + "sort": 0, + "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": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Disk", + "titleSize": "h6" + } + ], + "schemaVersion": 14, + "style": "dark", + "tags": [ + "kubernetes-mixin" + ], + "templating": { + "list": [ + { + "current": { + "text": "Prometheus", + "value": "Prometheus" + }, + "hide": 0, + "label": null, + "name": "datasource", + "options": [ + + ], + "query": "prometheus", + "refresh": 1, + "regex": "", + "type": "datasource" + }, + { + "allValue": null, + "current": { + "text": "prod", + "value": "prod" + }, + "datasource": "$datasource", + "hide": 2, + "includeAll": false, + "label": "cluster", + "multi": false, + "name": "cluster", + "options": [ + + ], + "query": "label_values(kube_node_info, cluster)", + "refresh": 1, + "regex": "", + "sort": 2, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "current": { + "text": "prod", + "value": "prod" + }, + "datasource": "$datasource", + "hide": 0, + "includeAll": false, + "label": "node", + "multi": false, + "name": "node", + "options": [ + + ], + "query": "label_values(kube_node_info{cluster=\"$cluster\"}, node)", + "refresh": 1, + "regex": "", + "sort": 2, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + } + ] + }, + "time": { + "from": "now-1h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "5s", + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ], + "time_options": [ + "5m", + "15m", + "1h", + "6h", + "12h", + "24h", + "2d", + "7d", + "30d" + ] + }, + "timezone": "", + "title": "Kubernetes / USE Method / Node", + "uid": "4ac4f123aae0ff6dbaf4f4f66120033b", + "version": 0 + } k8s-resources-cluster.json: |- { "annotations": { @@ -147,7 +2048,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) / sum(node:node_num_cpu:sum{cluster=\"$cluster\"})", + "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable_cpu_cores{cluster=\"$cluster\"})", "format": "time_series", "instant": true, "intervalFactor": 2, @@ -231,7 +2132,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\"}) / sum(node:node_num_cpu:sum{cluster=\"$cluster\"})", + "expr": "sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable_cpu_cores{cluster=\"$cluster\"})", "format": "time_series", "instant": true, "intervalFactor": 2, @@ -315,7 +2216,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "1 - sum(:node_memory_MemFreeCachedBuffers_bytes:sum{cluster=\"$cluster\"}) / sum(:node_memory_MemTotal_bytes:sum{cluster=\"$cluster\"})", + "expr": "1 - sum(:node_memory_MemFreeCachedBuffers_bytes:sum{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable_memory_bytes{cluster=\"$cluster\"})", "format": "time_series", "instant": true, "intervalFactor": 2, @@ -399,7 +2300,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) / sum(:node_memory_MemTotal_bytes:sum{cluster=\"$cluster\"})", + "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable_memory_bytes{cluster=\"$cluster\"})", "format": "time_series", "instant": true, "intervalFactor": 2, @@ -483,7 +2384,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\"}) / sum(:node_memory_MemTotal_bytes:sum{cluster=\"$cluster\"})", + "expr": "sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable_memory_bytes{cluster=\"$cluster\"})", "format": "time_series", "instant": true, "intervalFactor": 2, @@ -578,7 +2479,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{namespace}}", @@ -861,7 +2762,7 @@ data: "step": 10 }, { - "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -879,7 +2780,7 @@ data: "step": 10 }, { - "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -897,7 +2798,7 @@ data: "step": 10 }, { - "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -997,7 +2898,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(container_memory_rss{cluster=\"$cluster\", container_name!=\"\"}) by (namespace)", + "expr": "sum(container_memory_rss{cluster=\"$cluster\", container!=\"\"}) by (namespace)", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{namespace}}", @@ -1280,7 +3181,7 @@ data: "step": 10 }, { - "expr": "sum(container_memory_rss{cluster=\"$cluster\", container_name!=\"\"}) by (namespace)", + "expr": "sum(container_memory_rss{cluster=\"$cluster\", container!=\"\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -1298,7 +3199,7 @@ data: "step": 10 }, { - "expr": "sum(container_memory_rss{cluster=\"$cluster\", container_name!=\"\"}) by (namespace) / sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(container_memory_rss{cluster=\"$cluster\", container!=\"\"}) by (namespace) / sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -1316,7 +3217,7 @@ data: "step": 10 }, { - "expr": "sum(container_memory_rss{cluster=\"$cluster\", container_name!=\"\"}) by (namespace) / sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\"}) by (namespace)", + "expr": "sum(container_memory_rss{cluster=\"$cluster\", container!=\"\"}) by (namespace) / sum(kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\"}) by (namespace)", "format": "table", "instant": true, "intervalFactor": 2, @@ -1519,10 +3420,10 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod_name)", + "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{pod_name}}", + "legendFormat": "{{pod}}", "legendLink": null, "step": 10 } @@ -1748,7 +3649,7 @@ data: ], "targets": [ { - "expr": "sum(label_replace(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod)", + "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -1766,7 +3667,7 @@ data: "step": 10 }, { - "expr": "sum(label_replace(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", + "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -1784,7 +3685,7 @@ data: "step": 10 }, { - "expr": "sum(label_replace(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", + "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -1884,10 +3785,10 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"}) by (pod_name)", + "expr": "sum(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\"}) by (pod)", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{pod_name}}", + "legendFormat": "{{pod}}", "legendLink": null, "step": 10 } @@ -2167,7 +4068,7 @@ data: ], "targets": [ { - "expr": "sum(label_replace(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\",container_name!=\"\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod)", + "expr": "sum(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -2185,7 +4086,7 @@ data: "step": 10 }, { - "expr": "sum(label_replace(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\",container_name!=\"\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod) / sum(kube_pod_container_resource_requests_memory_bytes{namespace=\"$namespace\"}) by (pod)", + "expr": "sum(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\"}) by (pod) / sum(kube_pod_container_resource_requests_memory_bytes{namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -2203,7 +4104,7 @@ data: "step": 10 }, { - "expr": "sum(label_replace(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\",container_name!=\"\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod) / sum(kube_pod_container_resource_limits_memory_bytes{namespace=\"$namespace\"}) by (pod)", + "expr": "sum(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\"}) by (pod) / sum(kube_pod_container_resource_limits_memory_bytes{namespace=\"$namespace\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -2212,7 +4113,7 @@ data: "step": 10 }, { - "expr": "sum(label_replace(container_memory_rss{cluster=\"$cluster\", namespace=\"$namespace\",container_name!=\"\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod)", + "expr": "sum(container_memory_rss{cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -2221,7 +4122,7 @@ data: "step": 10 }, { - "expr": "sum(label_replace(container_memory_cache{cluster=\"$cluster\", namespace=\"$namespace\",container_name!=\"\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod)", + "expr": "sum(container_memory_cache{cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -2230,7 +4131,7 @@ data: "step": 10 }, { - "expr": "sum(label_replace(container_memory_swap{cluster=\"$cluster\", namespace=\"$namespace\",container_name!=\"\"}, \"pod\", \"$1\", \"pod_name\", \"(.*)\")) by (pod)", + "expr": "sum(container_memory_swap{cluster=\"$cluster\", namespace=\"$namespace\",container!=\"\"}) by (pod)", "format": "table", "instant": true, "intervalFactor": 2, @@ -2460,10 +4361,10 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"POD\", cluster=\"$cluster\"}) by (container_name)", + "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", cluster=\"$cluster\"}) by (container)", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{container_name}}", + "legendFormat": "{{container}}", "legendLink": null, "step": 10 } @@ -2689,7 +4590,7 @@ data: ], "targets": [ { - "expr": "sum(label_replace(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"POD\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container)", + "expr": "sum(namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, @@ -2707,7 +4608,7 @@ data: "step": 10 }, { - "expr": "sum(label_replace(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", + "expr": "sum(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)", "format": "table", "instant": true, "intervalFactor": 2, @@ -2725,7 +4626,7 @@ data: "step": 10 }, { - "expr": "sum(label_replace(namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container) / sum(kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)", + "expr": "sum(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)", "format": "table", "instant": true, "intervalFactor": 2, @@ -2825,26 +4726,26 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(container_memory_rss{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"POD\", container_name!=\"\"}) by (container_name)", + "expr": "sum(container_memory_rss{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", container!=\"\"}) by (container)", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{container_name}} (RSS)", + "legendFormat": "{{container}} (RSS)", "legendLink": null, "step": 10 }, { - "expr": "sum(container_memory_cache{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"POD\", container_name!=\"\"}) by (container_name)", + "expr": "sum(container_memory_cache{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", container!=\"\"}) by (container)", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{container_name}} (Cache)", + "legendFormat": "{{container}} (Cache)", "legendLink": null, "step": 10 }, { - "expr": "sum(container_memory_swap{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"POD\", container_name!=\"\"}) by (container_name)", + "expr": "sum(container_memory_swap{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", container!=\"\"}) by (container)", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{container_name}} (Swap)", + "legendFormat": "{{container}} (Swap)", "legendLink": null, "step": 10 } @@ -3124,7 +5025,7 @@ data: ], "targets": [ { - "expr": "sum(label_replace(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"POD\", container_name!=\"\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container)", + "expr": "sum(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", container!=\"\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, @@ -3142,7 +5043,7 @@ data: "step": 10 }, { - "expr": "sum(label_replace(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container) / sum(kube_pod_container_resource_requests_memory_bytes{namespace=\"$namespace\", pod=\"$pod\"}) by (container)", + "expr": "sum(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container) / sum(kube_pod_container_resource_requests_memory_bytes{namespace=\"$namespace\", pod=\"$pod\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, @@ -3160,7 +5061,7 @@ data: "step": 10 }, { - "expr": "sum(label_replace(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name!=\"\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container) / sum(kube_pod_container_resource_limits_memory_bytes{namespace=\"$namespace\", pod=\"$pod\"}) by (container)", + "expr": "sum(container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"\"}) by (container) / sum(kube_pod_container_resource_limits_memory_bytes{namespace=\"$namespace\", pod=\"$pod\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, @@ -3169,7 +5070,7 @@ data: "step": 10 }, { - "expr": "sum(label_replace(container_memory_rss{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name != \"\", container_name != \"POD\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container)", + "expr": "sum(container_memory_rss{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container != \"\", container != \"POD\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, @@ -3178,7 +5079,7 @@ data: "step": 10 }, { - "expr": "sum(label_replace(container_memory_cache{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name != \"\", container_name != \"POD\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container)", + "expr": "sum(container_memory_cache{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container != \"\", container != \"POD\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, @@ -3187,7 +5088,7 @@ data: "step": 10 }, { - "expr": "sum(label_replace(container_memory_swap{cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name != \"\", container_name != \"POD\"}, \"container\", \"$1\", \"container_name\", \"(.*)\")) by (container)", + "expr": "sum(container_memory_swap{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container != \"\", container != \"POD\"}) by (container)", "format": "table", "instant": true, "intervalFactor": 2, @@ -3444,7 +5345,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -3673,7 +5574,7 @@ data: ], "targets": [ { - "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -3682,7 +5583,7 @@ data: "step": 10 }, { - "expr": "sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "expr": "sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -3691,7 +5592,7 @@ data: "step": 10 }, { - "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -3700,7 +5601,7 @@ data: "step": 10 }, { - "expr": "sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "expr": "sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -3709,7 +5610,7 @@ data: "step": 10 }, { - "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -3809,7 +5710,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n ) by (pod)\n", + "expr": "sum(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{pod}}", @@ -4038,7 +5939,7 @@ data: ], "targets": [ { - "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n ) by (pod)\n", + "expr": "sum(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4047,7 +5948,7 @@ data: "step": 10 }, { - "expr": "sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "expr": "sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4056,7 +5957,7 @@ data: "step": 10 }, { - "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n ) by (pod)\n/sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "expr": "sum(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4065,7 +5966,7 @@ data: "step": 10 }, { - "expr": "sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "expr": "sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4074,7 +5975,7 @@ data: "step": 10 }, { - "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n ) by (pod)\n/sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", + "expr": "sum(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n/sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\", workload=\"$workload\", workload_type=\"$type\"}\n) by (pod)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4358,7 +6259,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{workload}} - {{workload_type}}", @@ -4632,7 +6533,7 @@ data: "step": 10 }, { - "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4641,7 +6542,7 @@ data: "step": 10 }, { - "expr": "sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4650,7 +6551,7 @@ data: "step": 10 }, { - "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4659,7 +6560,7 @@ data: "step": 10 }, { - "expr": "sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4668,7 +6569,7 @@ data: "step": 10 }, { - "expr": "sum(\n label_replace(\n namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_limits_cpu_cores{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -4768,7 +6669,7 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n ) by (workload, workload_type)\n", + "expr": "sum(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", "format": "time_series", "intervalFactor": 2, "legendFormat": "{{workload}} - {{workload_type}}", @@ -5042,7 +6943,7 @@ data: "step": 10 }, { - "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n ) by (workload, workload_type)\n", + "expr": "sum(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -5051,7 +6952,7 @@ data: "step": 10 }, { - "expr": "sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -5060,7 +6961,7 @@ data: "step": 10 }, { - "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n ) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -5069,7 +6970,7 @@ data: "step": 10 }, { - "expr": "sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", "format": "table", "instant": true, "intervalFactor": 2, @@ -5078,7 +6979,7 @@ data: "step": 10 }, { - "expr": "sum(\n label_replace(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container_name!=\"\"},\n \"pod\", \"$1\", \"pod_name\", \"(.*)\"\n ) * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n ) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n * on(namespace,pod) group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", + "expr": "sum(\n container_memory_usage_bytes{cluster=\"$cluster\", namespace=\"$namespace\", container!=\"\"}\n * on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n/sum(\n kube_pod_container_resource_limits_memory_bytes{cluster=\"$cluster\", namespace=\"$namespace\"}\n* on(namespace,pod)\n group_left(workload, workload_type) mixin_pod_workload{cluster=\"$cluster\", namespace=\"$namespace\"}\n) by (workload, workload_type)\n", "format": "table", "instant": true, "intervalFactor": 2, diff --git a/addons/grafana/dashboards-k8s.yaml b/addons/grafana/dashboards-k8s.yaml index d83a49cf..0f5a8c2e 100644 --- a/addons/grafana/dashboards-k8s.yaml +++ b/addons/grafana/dashboards-k8s.yaml @@ -4,1908 +4,7 @@ metadata: name: grafana-dashboards-k8s namespace: monitoring data: - k8s-cluster-rsrc-use.json: |- - { - "annotations": { - "list": [ - - ] - }, - "editable": true, - "gnetId": null, - "graphTooltip": 0, - "hideControls": false, - "links": [ - - ], - "refresh": "10s", - "rows": [ - { - "collapse": false, - "height": "250px", - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 10, - "id": 1, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "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": "node:cluster_cpu_utilisation:ratio{cluster=\"$cluster\"}", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{node}}", - "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", - "step": 10 - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "CPU Utilisation", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "percentunit", - "label": null, - "logBase": 1, - "max": 1, - "min": 0, - "show": true - }, - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": false - } - ] - }, - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 10, - "id": 2, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "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": "node:node_cpu_saturation_load1:{cluster=\"$cluster\"} / scalar(sum(min(kube_pod_info{cluster=\"$cluster\"}) by (node)))", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{node}}", - "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", - "step": 10 - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "CPU Saturation (Load1)", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "percentunit", - "label": null, - "logBase": 1, - "max": 1, - "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": "CPU", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 10, - "id": 3, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 0, - "links": [ - - ], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": true, - "steppedLine": false, - "targets": [ - { - "expr": "node:cluster_memory_utilisation:ratio{cluster=\"$cluster\"}", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{node}}", - "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", - "step": 10 - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Memory Utilisation", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "percentunit", - "label": null, - "logBase": 1, - "max": 1, - "min": 0, - "show": true - }, - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": false - } - ] - }, - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 10, - "id": 4, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 0, - "links": [ - - ], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": true, - "steppedLine": false, - "targets": [ - { - "expr": "node:node_memory_swap_io_bytes:sum_rate{cluster=\"$cluster\"}", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{node}}", - "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", - "step": 10 - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Memory Saturation (Swap I/O)", - "tooltip": { - "shared": false, - "sort": 0, - "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": "Memory", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 10, - "id": 5, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 0, - "links": [ - - ], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": true, - "steppedLine": false, - "targets": [ - { - "expr": "node:node_disk_utilisation:avg_irate{cluster=\"$cluster\"} / scalar(:kube_pod_info_node_count:{cluster=\"$cluster\"})", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{node}}", - "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", - "step": 10 - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Disk IO Utilisation", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "percentunit", - "label": null, - "logBase": 1, - "max": 1, - "min": 0, - "show": true - }, - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": false - } - ] - }, - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 10, - "id": 6, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "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": "node:node_disk_saturation:avg_irate{cluster=\"$cluster\"} / scalar(:kube_pod_info_node_count:{cluster=\"$cluster\"})", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{node}}", - "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", - "step": 10 - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Disk IO Saturation", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "percentunit", - "label": null, - "logBase": 1, - "max": 1, - "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": "Disk", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 10, - "id": 7, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 0, - "links": [ - - ], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": true, - "steppedLine": false, - "targets": [ - { - "expr": "node:node_net_utilisation:sum_irate{cluster=\"$cluster\"}", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{node}}", - "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", - "step": 10 - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Net Utilisation (Transmitted)", - "tooltip": { - "shared": false, - "sort": 0, - "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 - } - ] - }, - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 10, - "id": 8, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 0, - "links": [ - - ], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": true, - "steppedLine": false, - "targets": [ - { - "expr": "node:node_net_saturation:sum_irate{cluster=\"$cluster\"}", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{node}}", - "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", - "step": 10 - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Net Saturation (Dropped)", - "tooltip": { - "shared": false, - "sort": 0, - "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": "Network", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 10, - "id": 9, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 0, - "links": [ - - ], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 12, - "stack": true, - "steppedLine": false, - "targets": [ - { - "expr": "sum(max(node_filesystem_size_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\", cluster=\"$cluster\"} - node_filesystem_avail_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\", cluster=\"$cluster\"}) by (device,pod,namespace)) by (pod,namespace)\n/ scalar(sum(max(node_filesystem_size_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\", cluster=\"$cluster\"}) by (device,pod,namespace)))\n* on (namespace, pod) group_left (node) node_namespace_pod:kube_pod_info:{cluster=\"$cluster\"}\n", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{node}}", - "legendLink": "/d/4ac4f123aae0ff6dbaf4f4f66120033b/k8s-node-rsrc-use", - "step": 10 - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Disk Capacity", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "percentunit", - "label": null, - "logBase": 1, - "max": 1, - "min": 0, - "show": true - }, - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": false - } - ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Storage", - "titleSize": "h6" - } - ], - "schemaVersion": 14, - "style": "dark", - "tags": [ - "kubernetes-mixin" - ], - "templating": { - "list": [ - { - "current": { - "text": "Prometheus", - "value": "Prometheus" - }, - "hide": 0, - "label": null, - "name": "datasource", - "options": [ - - ], - "query": "prometheus", - "refresh": 1, - "regex": "", - "type": "datasource" - }, - { - "allValue": null, - "current": { - "text": "prod", - "value": "prod" - }, - "datasource": "$datasource", - "hide": 2, - "includeAll": false, - "label": "cluster", - "multi": false, - "name": "cluster", - "options": [ - - ], - "query": "label_values(kube_node_info, cluster)", - "refresh": 1, - "regex": "", - "sort": 2, - "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": "", - "title": "Kubernetes / USE Method / Cluster", - "uid": "a6e7d1362e1ddbb79db21d5bb40d7137", - "version": 0 - } - k8s-node-rsrc-use.json: |- - { - "annotations": { - "list": [ - - ] - }, - "editable": true, - "gnetId": null, - "graphTooltip": 0, - "hideControls": false, - "links": [ - - ], - "refresh": "10s", - "rows": [ - { - "collapse": false, - "height": "250px", - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "id": 1, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "node:node_cpu_utilisation:avg1m{cluster=\"$cluster\", node=\"$node\"}", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "Utilisation", - "legendLink": null, - "step": 10 - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "CPU Utilisation", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "percentunit", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - }, - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": false - } - ] - }, - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "id": 2, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "node:node_cpu_saturation_load1:{cluster=\"$cluster\", node=\"$node\"}", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "Saturation", - "legendLink": null, - "step": 10 - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "CPU Saturation (Load1)", - "tooltip": { - "shared": false, - "sort": 0, - "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": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": false - } - ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "CPU", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "id": 3, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "node:node_memory_utilisation:{cluster=\"$cluster\", node=\"$node\"}", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "Memory", - "legendLink": null, - "step": 10 - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Memory Utilisation", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "percentunit", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - }, - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": false - } - ] - }, - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "id": 4, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "node:node_memory_swap_io_bytes:sum_rate{cluster=\"$cluster\", node=\"$node\"}", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "Swap IO", - "legendLink": null, - "step": 10 - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Memory Saturation (Swap I/O)", - "tooltip": { - "shared": false, - "sort": 0, - "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": "Memory", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "id": 5, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "node:node_disk_utilisation:avg_irate{cluster=\"$cluster\", node=\"$node\"}", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "Utilisation", - "legendLink": null, - "step": 10 - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Disk IO Utilisation", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "percentunit", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - }, - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": false - } - ] - }, - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "id": 6, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "node:node_disk_saturation:avg_irate{cluster=\"$cluster\", node=\"$node\"}", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "Saturation", - "legendLink": null, - "step": 10 - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Disk IO Saturation", - "tooltip": { - "shared": false, - "sort": 0, - "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": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": false - } - ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Disk", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "id": 7, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "node:node_net_utilisation:sum_irate{cluster=\"$cluster\", node=\"$node\"}", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "Utilisation", - "legendLink": null, - "step": 10 - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Net Utilisation (Transmitted)", - "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" - }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ - { - "format": "Bps", - "label": null, - "logBase": 1, - "max": null, - "min": 0, - "show": true - }, - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": false - } - ] - }, - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "id": 8, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "node:node_net_saturation:sum_irate{cluster=\"$cluster\", node=\"$node\"}", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "Saturation", - "legendLink": null, - "step": 10 - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Net Saturation (Dropped)", - "tooltip": { - "shared": false, - "sort": 0, - "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": "Net", - "titleSize": "h6" - }, - { - "collapse": false, - "height": "250px", - "panels": [ - { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, - "datasource": "$datasource", - "fill": 1, - "id": 9, - "legend": { - "avg": false, - "current": false, - "max": false, - "min": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null as zero", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 12, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "node:node_filesystem_usage:{cluster=\"$cluster\"}\n* on (namespace, pod) group_left (node) node_namespace_pod:kube_pod_info:{cluster=\"$cluster\", node=\"$node\"}\n", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "{{device}}", - "legendLink": null, - "step": 10 - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Disk Utilisation", - "tooltip": { - "shared": false, - "sort": 0, - "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": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": false - } - ] - } - ], - "repeat": null, - "repeatIteration": null, - "repeatRowId": null, - "showTitle": true, - "title": "Disk", - "titleSize": "h6" - } - ], - "schemaVersion": 14, - "style": "dark", - "tags": [ - "kubernetes-mixin" - ], - "templating": { - "list": [ - { - "current": { - "text": "Prometheus", - "value": "Prometheus" - }, - "hide": 0, - "label": null, - "name": "datasource", - "options": [ - - ], - "query": "prometheus", - "refresh": 1, - "regex": "", - "type": "datasource" - }, - { - "allValue": null, - "current": { - "text": "prod", - "value": "prod" - }, - "datasource": "$datasource", - "hide": 2, - "includeAll": false, - "label": "cluster", - "multi": false, - "name": "cluster", - "options": [ - - ], - "query": "label_values(kube_node_info, cluster)", - "refresh": 1, - "regex": "", - "sort": 2, - "tagValuesQuery": "", - "tags": [ - - ], - "tagsQuery": "", - "type": "query", - "useTags": false - }, - { - "allValue": null, - "current": { - "text": "prod", - "value": "prod" - }, - "datasource": "$datasource", - "hide": 0, - "includeAll": false, - "label": "node", - "multi": false, - "name": "node", - "options": [ - - ], - "query": "label_values(kube_node_info{cluster=\"$cluster\"}, node)", - "refresh": 1, - "regex": "", - "sort": 2, - "tagValuesQuery": "", - "tags": [ - - ], - "tagsQuery": "", - "type": "query", - "useTags": false - } - ] - }, - "time": { - "from": "now-1h", - "to": "now" - }, - "timepicker": { - "refresh_intervals": [ - "5s", - "10s", - "30s", - "1m", - "5m", - "15m", - "30m", - "1h", - "2h", - "1d" - ], - "time_options": [ - "5m", - "15m", - "1h", - "6h", - "12h", - "24h", - "2d", - "7d", - "30d" - ] - }, - "timezone": "", - "title": "Kubernetes / USE Method / Node", - "uid": "4ac4f123aae0ff6dbaf4f4f66120033b", - "version": 0 - } - nodes.json: |- + apiserver.json: |- { "__inputs": [ @@ -1933,116 +32,88 @@ data: "collapsed": false, "panels": [ { - "aliasColors": { - - }, - "bars": false, - "dashLength": 10, - "dashes": false, + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": [ + "#299c46", + "rgba(237, 129, 40, 0.89)", + "#d44a3a" + ], "datasource": "$datasource", - "fill": 1, + "format": "none", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, "gridPos": { }, "id": 2, - "legend": { - "alignAsTable": false, - "avg": false, - "current": false, - "max": false, - "min": false, - "rightSide": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, + "interval": null, "links": [ ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } ], - "spaceLength": 10, - "span": 6, - "stack": false, - "steppedLine": false, + "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": "max(node_load1{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", + "expr": "sum(up{job=\"apiserver\"})", "format": "time_series", "intervalFactor": 2, - "legendFormat": "load 1m", + "legendFormat": "", "refId": "A" - }, - { - "expr": "max(node_load5{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "load 5m", - "refId": "B" - }, - { - "expr": "max(node_load15{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "load 15m", - "refId": "C" - }, - { - "expr": "count(node_cpu_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\", mode=\"user\"})", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "logical cores", - "refId": "D" } ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "System load", + "thresholds": "", + "title": "Up", "tooltip": { - "shared": false, - "sort": 0, - "value_type": "individual" + "shared": false }, - "type": "graph", - "xaxis": { - "buckets": null, - "mode": "time", - "name": null, - "show": true, - "values": [ - - ] - }, - "yaxes": [ + "type": "singlestat", + "valueFontSize": "80%", + "valueMaps": [ { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": true - }, - { - "format": "short", - "label": null, - "logBase": 1, - "max": null, - "min": null, - "show": true + "op": "=", + "text": "N/A", + "value": "null" } - ] + ], + "valueName": "min" }, { "aliasColors": { @@ -2083,16 +154,37 @@ data: ], "spaceLength": 10, - "span": 6, + "span": 5, "stack": false, "steppedLine": false, "targets": [ { - "expr": "sum by (cpu) (irate(node_cpu_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", mode!=\"idle\", instance=\"$instance\"}[5m]))", + "expr": "sum(rate(apiserver_request_total{job=\"apiserver\", instance=~\"$instance\",code=~\"2..\"}[5m]))", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{cpu}}", + "legendFormat": "2xx", "refId": "A" + }, + { + "expr": "sum(rate(apiserver_request_total{job=\"apiserver\", instance=~\"$instance\",code=~\"3..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "3xx", + "refId": "B" + }, + { + "expr": "sum(rate(apiserver_request_total{job=\"apiserver\", instance=~\"$instance\",code=~\"4..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "4xx", + "refId": "C" + }, + { + "expr": "sum(rate(apiserver_request_total{job=\"apiserver\", instance=~\"$instance\",code=~\"5..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "5xx", + "refId": "D" } ], "thresholds": [ @@ -2100,7 +192,7 @@ data: ], "timeFrom": null, "timeShift": null, - "title": "Usage Per Core", + "title": "RPC Rate", "tooltip": { "shared": false, "sort": 0, @@ -2118,7 +210,7 @@ data: }, "yaxes": [ { - "format": "percentunit", + "format": "ops", "label": null, "logBase": 1, "max": null, @@ -2126,7 +218,98 @@ data: "show": true }, { - "format": "percentunit", + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 4, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "total": false, + "values": "true" + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 5, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.99, sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\", instance=~\"$instance\"}[5m])) by (verb, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{verb}}", + "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": null, + "show": true + }, + { + "format": "s", "label": null, "logBase": 1, "max": null, @@ -2160,17 +343,17 @@ data: "gridPos": { }, - "id": 4, + "id": 5, "legend": { - "alignAsTable": "true", - "avg": "true", - "current": "true", - "max": "false", - "min": "false", - "rightSide": "true", - "show": "true", - "total": "false", - "values": "true" + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": false, + "total": false, + "values": false }, "lines": true, "linewidth": 1, @@ -2187,15 +370,15 @@ data: ], "spaceLength": 10, - "span": 9, + "span": 6, "stack": false, "steppedLine": false, "targets": [ { - "expr": "max (sum by (cpu) (irate(node_cpu_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", mode!=\"idle\", instance=\"$instance\"}[2m])) ) * 100\n", + "expr": "sum(rate(workqueue_adds_total{job=\"apiserver\", instance=~\"$instance\"}[5m])) by (instance, name)", "format": "time_series", - "intervalFactor": 10, - "legendFormat": "{{ cpu }}", + "intervalFactor": 2, + "legendFormat": "{{instance}} {{name}}", "refId": "A" } ], @@ -2204,7 +387,7 @@ data: ], "timeFrom": null, "timeShift": null, - "title": "CPU Utilization", + "title": "Work Queue Add Rate", "tooltip": { "shared": false, "sort": 0, @@ -2222,120 +405,23 @@ data: }, "yaxes": [ { - "format": "percent", + "format": "ops", "label": null, "logBase": 1, - "max": 100, + "max": null, "min": 0, "show": true }, { - "format": "percent", + "format": "ops", "label": null, "logBase": 1, - "max": 100, + "max": null, "min": 0, "show": true } ] }, - { - "cacheTimeout": null, - "colorBackground": false, - "colorValue": false, - "colors": [ - "rgba(50, 172, 45, 0.97)", - "rgba(237, 129, 40, 0.89)", - "rgba(245, 54, 54, 0.9)" - ], - "datasource": "$datasource", - "format": "percent", - "gauge": { - "maxValue": 100, - "minValue": 0, - "show": true, - "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": "avg(sum by (cpu) (irate(node_cpu_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", mode!=\"idle\", instance=\"$instance\"}[2m]))) * 100\n", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "", - "refId": "A" - } - ], - "thresholds": "80, 90", - "title": "CPU Usage", - "tooltip": { - "shared": false - }, - "type": "singlestat", - "valueFontSize": "80%", - "valueMaps": [ - { - "op": "=", - "text": "N/A", - "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": { @@ -2356,7 +442,7 @@ data: "max": false, "min": false, "rightSide": false, - "show": true, + "show": false, "total": false, "values": false }, @@ -2375,37 +461,16 @@ data: ], "spaceLength": 10, - "span": 9, + "span": 6, "stack": false, "steppedLine": false, "targets": [ { - "expr": "max(\n node_memory_MemTotal_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_MemFree_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_Buffers_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_Cached_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n)\n", + "expr": "sum(rate(workqueue_depth{job=\"apiserver\", instance=~\"$instance\"}[5m])) by (instance, name)", "format": "time_series", "intervalFactor": 2, - "legendFormat": "memory used", + "legendFormat": "{{instance}} {{name}}", "refId": "A" - }, - { - "expr": "max(node_memory_Buffers_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "memory buffers", - "refId": "B" - }, - { - "expr": "max(node_memory_Cached_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "memory cached", - "refId": "C" - }, - { - "expr": "max(node_memory_MemFree_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "memory free", - "refId": "D" } ], "thresholds": [ @@ -2413,7 +478,7 @@ data: ], "timeFrom": null, "timeShift": null, - "title": "Memory Usage", + "title": "Work Queue Depth", "tooltip": { "shared": false, "sort": 0, @@ -2431,7 +496,98 @@ data: }, "yaxes": [ { - "format": "bytes", + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 7, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "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(workqueue_queue_duration_seconds_bucket{job=\"apiserver\", instance=~\"$instance\"}[5m])) by (instance, name, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} {{name}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Work Queue Latency", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "s", "label": null, "logBase": 1, "max": null, @@ -2439,7 +595,7 @@ data: "show": true }, { - "format": "bytes", + "format": "s", "label": null, "logBase": 1, "max": null, @@ -2447,90 +603,6 @@ data: "show": true } ] - }, - { - "cacheTimeout": null, - "colorBackground": false, - "colorValue": false, - "colors": [ - "rgba(50, 172, 45, 0.97)", - "rgba(237, 129, 40, 0.89)", - "rgba(245, 54, 54, 0.9)" - ], - "datasource": "$datasource", - "format": "percent", - "gauge": { - "maxValue": 100, - "minValue": 0, - "show": true, - "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(\n (\n (\n node_memory_MemTotal_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_MemFree_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_Buffers_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_memory_Cached_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n )\n / node_memory_MemTotal_bytes{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n ) * 100)\n", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "", - "refId": "A" - } - ], - "thresholds": "80, 90", - "title": "Memory Usage", - "tooltip": { - "shared": false - }, - "type": "singlestat", - "valueFontSize": "80%", - "valueMaps": [ - { - "op": "=", - "text": "N/A", - "value": "null" - } - ], - "valueName": "current" } ], "repeat": null, @@ -2581,40 +653,19 @@ data: "renderer": "flot", "repeat": null, "seriesOverrides": [ - { - "alias": "read", - "yaxis": 1 - }, - { - "alias": "io time", - "yaxis": 2 - } + ], "spaceLength": 10, - "span": 6, + "span": 4, "stack": false, "steppedLine": false, "targets": [ { - "expr": "max(rate(node_disk_read_bytes_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}[2m]))", + "expr": "etcd_helper_cache_entry_total{job=\"apiserver\", instance=~\"$instance\"}", "format": "time_series", "intervalFactor": 2, - "legendFormat": "read", + "legendFormat": "{{instance}}", "refId": "A" - }, - { - "expr": "max(rate(node_disk_written_bytes_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}[2m]))", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "written", - "refId": "B" - }, - { - "expr": "max(rate(node_disk_io_time_seconds_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}[2m]))", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "io time", - "refId": "C" } ], "thresholds": [ @@ -2622,7 +673,7 @@ data: ], "timeFrom": null, "timeShift": null, - "title": "Disk I/O", + "title": "ETCD Cache Entry Total", "tooltip": { "shared": false, "sort": 0, @@ -2640,19 +691,19 @@ data: }, "yaxes": [ { - "format": "bytes", + "format": "short", "label": null, "logBase": 1, "max": null, - "min": null, + "min": 0, "show": true }, { - "format": "ms", + "format": "short", "label": null, "logBase": 1, "max": null, - "min": null, + "min": 0, "show": true } ] @@ -2696,16 +747,23 @@ data: ], "spaceLength": 10, - "span": 6, + "span": 4, "stack": false, "steppedLine": false, "targets": [ { - "expr": "node:node_filesystem_usage:{cluster=\"$cluster\"}", + "expr": "sum(rate(etcd_helper_cache_hit_total{job=\"apiserver\",instance=~\"$instance\"}[5m])) by (intance)", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{device}}", + "legendFormat": "{{instance}} hit", "refId": "A" + }, + { + "expr": "sum(rate(etcd_helper_cache_miss_total{job=\"apiserver\",instance=~\"$instance\"}[5m])) by (instance)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} miss", + "refId": "B" } ], "thresholds": [ @@ -2713,7 +771,7 @@ data: ], "timeFrom": null, "timeShift": null, - "title": "Disk Space Usage", + "title": "ETCD Cache Hit/Miss Rate", "tooltip": { "shared": false, "sort": 0, @@ -2731,36 +789,23 @@ data: }, "yaxes": [ { - "format": "percentunit", + "format": "ops", "label": null, "logBase": 1, "max": null, - "min": null, + "min": 0, "show": true }, { - "format": "percentunit", + "format": "ops", "label": null, "logBase": 1, "max": null, - "min": 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": { @@ -2800,15 +845,126 @@ data: ], "spaceLength": 10, - "span": 6, + "span": 4, "stack": false, "steppedLine": false, "targets": [ { - "expr": "max(rate(node_network_receive_bytes_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\", device!~\"lo\"}[5m]))", + "expr": "histogram_quantile(0.99,sum(rate(etcd_request_cache_get_duration_seconds_bucket{job=\"apiserver\",instance=~\"$instance\"}[5m])) by (instance, le))", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{device}}", + "legendFormat": "{{instance}} get", + "refId": "A" + }, + { + "expr": "histogram_quantile(0.99,sum(rate(etcd_request_cache_add_duration_seconds_bucket{job=\"apiserver\",instance=~\"$instance\"}[5m])) by (instance, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} miss", + "refId": "B" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "ETCD Cache 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": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 11, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 4, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "process_resident_memory_bytes{job=\"apiserver\",instance=~\"$instance\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}}", "refId": "A" } ], @@ -2817,7 +973,7 @@ data: ], "timeFrom": null, "timeShift": null, - "title": "Network Received", + "title": "Memory", "tooltip": { "shared": false, "sort": 0, @@ -2863,6 +1019,1208 @@ data: "fill": 1, "gridPos": { + }, + "id": 12, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "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{job=\"apiserver\",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, + "gridPos": { + + }, + "id": 13, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "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{job=\"apiserver\",instance=~\"$instance\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Goroutines", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Dashboard Row", + "titleSize": "h6", + "type": "row" + } + ], + "schemaVersion": 14, + "style": "dark", + "tags": [ + "kubernetes-mixin" + ], + "templating": { + "list": [ + { + "current": { + "text": "Prometheus", + "value": "Prometheus" + }, + "hide": 0, + "label": null, + "name": "datasource", + "options": [ + + ], + "query": "prometheus", + "refresh": 1, + "regex": "", + "type": "datasource" + }, + { + "allValue": null, + "current": { + + }, + "datasource": "$datasource", + "hide": 0, + "includeAll": true, + "label": null, + "multi": false, + "name": "instance", + "options": [ + + ], + "query": "label_values(apiserver_request_total{job=\"apiserver\"}, instance)", + "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": "", + "title": "Kubernetes / API server", + "uid": "09ec8aa1e996d6ffcd6817bbaff4db1b", + "version": 0 + } + controller-manager.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": "$datasource", + "format": "none", + "gauge": { + "maxValue": 100, + "minValue": 0, + "show": false, + "thresholdLabels": false, + "thresholdMarkers": true + }, + "gridPos": { + + }, + "id": 2, + "interval": null, + "links": [ + + ], + "mappingType": 1, + "mappingTypes": [ + { + "name": "value to text", + "value": 1 + }, + { + "name": "range to text", + "value": 2 + } + ], + "maxDataPoints": 100, + "nullPointMode": "connected", + "nullText": null, + "postfix": "", + "postfixFontSize": "50%", + "prefix": "", + "prefixFontSize": "50%", + "rangeMaps": [ + { + "from": "null", + "text": "N/A", + "to": "null" + } + ], + "span": 2, + "sparkline": { + "fillColor": "rgba(31, 118, 189, 0.18)", + "full": false, + "lineColor": "rgb(31, 120, 193)", + "show": false + }, + "tableColumn": "", + "targets": [ + { + "expr": "sum(up{job=\"kube-controller-manager\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "", + "refId": "A" + } + ], + "thresholds": "", + "title": "Up", + "tooltip": { + "shared": false + }, + "type": "singlestat", + "valueFontSize": "80%", + "valueMaps": [ + { + "op": "=", + "text": "N/A", + "value": "null" + } + ], + "valueName": "min" + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 3, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "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": 10, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum(rate(workqueue_adds_total{job=\"kube-controller-manager\", instance=~\"$instance\"}[5m])) by (instance, name)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} {{name}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Work Queue Add Rate", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Dashboard Row", + "titleSize": "h6", + "type": "row" + }, + { + "collapse": false, + "collapsed": false, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 4, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "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": "sum(rate(workqueue_depth{job=\"kube-controller-manager\", instance=~\"$instance\"}[5m])) by (instance, name)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} {{name}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Work Queue Depth", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Dashboard Row", + "titleSize": "h6", + "type": "row" + }, + { + "collapse": false, + "collapsed": false, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 5, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "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(workqueue_queue_duration_seconds_bucket{job=\"kube-controller-manager\", instance=~\"$instance\"}[5m])) by (instance, name, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} {{name}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Work Queue Latency", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Dashboard Row", + "titleSize": "h6", + "type": "row" + }, + { + "collapse": false, + "collapsed": false, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 6, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "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": "sum(rate(rest_client_requests_total{job=\"kube-controller-manager\", instance=~\"$instance\",code=~\"2..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "2xx", + "refId": "A" + }, + { + "expr": "sum(rate(rest_client_requests_total{job=\"kube-controller-manager\", instance=~\"$instance\",code=~\"3..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "3xx", + "refId": "B" + }, + { + "expr": "sum(rate(rest_client_requests_total{job=\"kube-controller-manager\", instance=~\"$instance\",code=~\"4..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "4xx", + "refId": "C" + }, + { + "expr": "sum(rate(rest_client_requests_total{job=\"kube-controller-manager\", instance=~\"$instance\",code=~\"5..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "5xx", + "refId": "D" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Kube API Request 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": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 7, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 8, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_latency_seconds_bucket{job=\"kube-controller-manager\", instance=~\"$instance\", verb=\"POST\"}[5m])) by (verb, url, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{verb}} {{url}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Post Request Latency 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": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 8, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "total": false, + "values": "true" + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_latency_seconds_bucket{job=\"kube-controller-manager\", instance=~\"$instance\", verb=\"GET\"}[5m])) by (verb, url, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{verb}} {{url}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Get Request Latency 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": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 9, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 4, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "process_resident_memory_bytes{job=\"kube-controller-manager\",instance=~\"$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": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 10, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 4, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "rate(process_cpu_seconds_total{job=\"kube-controller-manager\",instance=~\"$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": "$datasource", + "fill": 1, + "gridPos": { + }, "id": 11, "legend": { @@ -2891,15 +2249,15 @@ data: ], "spaceLength": 10, - "span": 6, + "span": 4, "stack": false, "steppedLine": false, "targets": [ { - "expr": "max(rate(node_network_transmit_bytes_total{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\", device!~\"lo\"}[5m]))", + "expr": "go_goroutines{job=\"kube-controller-manager\",instance=~\"$instance\"}", "format": "time_series", "intervalFactor": 2, - "legendFormat": "{{device}}", + "legendFormat": "{{instance}}", "refId": "A" } ], @@ -2908,118 +2266,7 @@ data: ], "timeFrom": null, "timeShift": null, - "title": "Network Transmitted", - "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 - } - ] - } - ], - "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": "$datasource", - "fill": 1, - "gridPos": { - - }, - "id": 12, - "legend": { - "alignAsTable": false, - "avg": false, - "current": false, - "max": false, - "min": false, - "rightSide": false, - "show": true, - "total": false, - "values": false - }, - "lines": true, - "linewidth": 1, - "links": [ - - ], - "nullPointMode": "null", - "percentage": false, - "pointradius": 5, - "points": false, - "renderer": "flot", - "repeat": null, - "seriesOverrides": [ - - ], - "spaceLength": 10, - "span": 9, - "stack": false, - "steppedLine": false, - "targets": [ - { - "expr": "max(\n node_filesystem_files{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_filesystem_files_free{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n)\n", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "inodes used", - "refId": "A" - }, - { - "expr": "max(node_filesystem_files_free{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"})", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "inodes free", - "refId": "B" - } - ], - "thresholds": [ - - ], - "timeFrom": null, - "timeShift": null, - "title": "Inodes Usage", + "title": "Goroutines", "tooltip": { "shared": false, "sort": 0, @@ -3053,90 +2300,6 @@ data: "show": true } ] - }, - { - "cacheTimeout": null, - "colorBackground": false, - "colorValue": false, - "colors": [ - "rgba(50, 172, 45, 0.97)", - "rgba(237, 129, 40, 0.89)", - "rgba(245, 54, 54, 0.9)" - ], - "datasource": "$datasource", - "format": "percent", - "gauge": { - "maxValue": 100, - "minValue": 0, - "show": true, - "thresholdLabels": false, - "thresholdMarkers": true - }, - "gridPos": { - - }, - "id": 13, - "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(\n (\n (\n node_filesystem_files{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n - node_filesystem_files_free{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n )\n / node_filesystem_files{cluster=\"$cluster\", job=\"node-exporter\", instance=\"$instance\"}\n ) * 100)\n", - "format": "time_series", - "intervalFactor": 2, - "legendFormat": "", - "refId": "A" - } - ], - "thresholds": "80, 90", - "title": "Inodes Usage", - "tooltip": { - "shared": false - }, - "type": "singlestat", - "valueFontSize": "80%", - "valueMaps": [ - { - "op": "=", - "text": "N/A", - "value": "null" - } - ], - "valueName": "current" } ], "repeat": null, @@ -3175,43 +2338,17 @@ data: "allValue": null, "current": { - }, - "datasource": "$datasource", - "hide": 2, - "includeAll": false, - "label": "cluster", - "multi": false, - "name": "cluster", - "options": [ - - ], - "query": "label_values(kube_pod_info, cluster)", - "refresh": 2, - "regex": "", - "sort": 0, - "tagValuesQuery": "", - "tags": [ - - ], - "tagsQuery": "", - "type": "query", - "useTags": false - }, - { - "allValue": null, - "current": { - }, "datasource": "$datasource", "hide": 0, - "includeAll": false, + "includeAll": true, "label": null, "multi": false, "name": "instance", "options": [ ], - "query": "label_values(node_boot_time_seconds{cluster=\"$cluster\", job=\"node-exporter\"}, instance)", + "query": "label_values(process_cpu_seconds_total{job=\"kube-controller-manager\"}, instance)", "refresh": 2, "regex": "", "sort": 0, @@ -3255,8 +2392,8 @@ data: ] }, "timezone": "", - "title": "Kubernetes / Nodes", - "uid": "fa49a4706d07a042595b664c87fb33ea", + "title": "Kubernetes / Controller Manager", + "uid": "72e0e05bef5099e5f049b05fdc429ed4", "version": 0 } persistentvolumesusage.json: |- @@ -3894,10 +3031,10 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum by(container_name) (container_memory_usage_bytes{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\", container_name=~\"$container\", container_name!=\"POD\"})", + "expr": "sum by(container) (container_memory_usage_bytes{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container=~\"$container\", container!=\"POD\"})", "format": "time_series", "intervalFactor": 2, - "legendFormat": "Current: {{ container_name }}", + "legendFormat": "Current: {{ container }}", "refId": "A" }, { @@ -3915,10 +3052,10 @@ data: "refId": "C" }, { - "expr": "sum by(container_name) (container_memory_cache{job=\"kubernetes-cadvisor\", namespace=\"$namespace\", pod_name=~\"$pod\", container_name=~\"$container\", container_name!=\"POD\"})", + "expr": "sum by(container) (container_memory_cache{job=\"kubernetes-cadvisor\", namespace=\"$namespace\", pod=~\"$pod\", container=~\"$container\", container!=\"POD\"})", "format": "time_series", "intervalFactor": 2, - "legendFormat": "Cache: {{ container_name }}", + "legendFormat": "Cache: {{ container }}", "refId": "D" } ], @@ -4019,10 +3156,10 @@ data: "steppedLine": false, "targets": [ { - "expr": "sum by (container_name) (rate(container_cpu_usage_seconds_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", image!=\"\", pod_name=\"$pod\", container_name=~\"$container\", container_name!=\"POD\"}[1m]))", + "expr": "sum by (container) (irate(container_cpu_usage_seconds_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", image!=\"\", pod=\"$pod\", container=~\"$container\", container!=\"POD\"}[4m]))", "format": "time_series", "intervalFactor": 2, - "legendFormat": "Current: {{ container_name }}", + "legendFormat": "Current: {{ container }}", "refId": "A" }, { @@ -4137,17 +3274,17 @@ data: "steppedLine": false, "targets": [ { - "expr": "sort_desc(sum by (pod_name) (rate(container_network_receive_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\"}[1m])))", + "expr": "sort_desc(sum by (pod) (irate(container_network_receive_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}[4m])))", "format": "time_series", "intervalFactor": 2, - "legendFormat": "RX: {{ pod_name }}", + "legendFormat": "RX: {{ pod }}", "refId": "A" }, { - "expr": "sort_desc(sum by (pod_name) (rate(container_network_transmit_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod_name=\"$pod\"}[1m])))", + "expr": "sort_desc(sum by (pod) (irate(container_network_transmit_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}[4m])))", "format": "time_series", "intervalFactor": 2, - "legendFormat": "TX: {{ pod_name }}", + "legendFormat": "TX: {{ pod }}", "refId": "B" } ], @@ -4468,6 +3605,1042 @@ data: "uid": "ab4f13a9892a76a4d21ce8c2445bf4ea", "version": 0 } + scheduler.json: |- + { + "__inputs": [ + + ], + "__requires": [ + + ], + "annotations": { + "list": [ + + ] + }, + "editable": false, + "gnetId": null, + "graphTooltip": 0, + "hideControls": false, + "id": null, + "links": [ + + ], + "refresh": "", + "rows": [ + { + "collapse": false, + "collapsed": false, + "panels": [ + { + "cacheTimeout": null, + "colorBackground": false, + "colorValue": false, + "colors": [ + "#299c46", + "rgba(237, 129, 40, 0.89)", + "#d44a3a" + ], + "datasource": "$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": "", + "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{job=\"kube-scheduler\"})", + "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" + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 3, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "total": false, + "values": "true" + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 5, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum(rate(scheduler_e2e_scheduling_duration_seconds_count{job=\"kube-scheduler\", instance=~\"$instance\"}[5m])) by (instance)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} e2e", + "refId": "A" + }, + { + "expr": "sum(rate(scheduler_binding_duration_seconds_count{job=\"kube-scheduler\", instance=~\"$instance\"}[5m])) by (instance)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} binding", + "refId": "B" + }, + { + "expr": "sum(rate(scheduler_scheduling_algorithm_duration_seconds_count{job=\"kube-scheduler\", instance=~\"$instance\"}[5m])) by (instance)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} scheduling algorithm", + "refId": "C" + }, + { + "expr": "sum(rate(scheduler_volume_scheduling_duration_seconds_count{job=\"kube-scheduler\", instance=~\"$instance\"}[5m])) by (instance)", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} volume", + "refId": "D" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Scheduling Rate", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 4, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "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": 5, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.99, sum(rate(scheduler_e2e_scheduling_duration_seconds_bucket{job=\"kube-scheduler\",instance=~\"$instance\"}[5m])) by (instance, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} e2e", + "refId": "A" + }, + { + "expr": "histogram_quantile(0.99, sum(rate(scheduler_binding_duration_seconds_bucket{job=\"kube-scheduler\",instance=~\"$instance\"}[5m])) by (instance, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} binding", + "refId": "B" + }, + { + "expr": "histogram_quantile(0.99, sum(rate(scheduler_scheduling_algorithm_duration_seconds_bucket{job=\"kube-scheduler\",instance=~\"$instance\"}[5m])) by (instance, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} scheduling algorithm", + "refId": "C" + }, + { + "expr": "histogram_quantile(0.99, sum(rate(scheduler_volume_scheduling_duration_seconds_bucket{job=\"kube-scheduler\",instance=~\"$instance\"}[5m])) by (instance, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}} volume", + "refId": "D" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Scheduling latency 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": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 5, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "repeat": null, + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 4, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum(rate(rest_client_requests_total{job=\"kube-scheduler\", instance=~\"$instance\",code=~\"2..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "2xx", + "refId": "A" + }, + { + "expr": "sum(rate(rest_client_requests_total{job=\"kube-scheduler\", instance=~\"$instance\",code=~\"3..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "3xx", + "refId": "B" + }, + { + "expr": "sum(rate(rest_client_requests_total{job=\"kube-scheduler\", instance=~\"$instance\",code=~\"4..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "4xx", + "refId": "C" + }, + { + "expr": "sum(rate(rest_client_requests_total{job=\"kube-scheduler\", instance=~\"$instance\",code=~\"5..\"}[5m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "5xx", + "refId": "D" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Kube API Request Rate", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "ops", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 6, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "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": 8, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_latency_seconds_bucket{job=\"kube-scheduler\", instance=~\"$instance\", verb=\"POST\"}[5m])) by (verb, url, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{verb}} {{url}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Post Request Latency 99th Quantile", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Dashboard Row", + "titleSize": "h6", + "type": "row" + }, + { + "collapse": false, + "collapsed": false, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 7, + "legend": { + "alignAsTable": "true", + "avg": false, + "current": "true", + "max": false, + "min": false, + "rightSide": "true", + "show": "true", + "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_latency_seconds_bucket{job=\"kube-scheduler\", instance=~\"$instance\", verb=\"GET\"}[5m])) by (verb, url, le))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{verb}} {{url}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Get Request Latency 99th Quantile", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Dashboard Row", + "titleSize": "h6", + "type": "row" + }, + { + "collapse": false, + "collapsed": false, + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 8, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "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{job=\"kube-scheduler\", 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, + "gridPos": { + + }, + "id": 9, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "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{job=\"kube-scheduler\", instance=~\"$instance\"}[5m])", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "CPU usage", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "bytes", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "bytes", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "gridPos": { + + }, + "id": 10, + "legend": { + "alignAsTable": false, + "avg": false, + "current": false, + "max": false, + "min": false, + "rightSide": false, + "show": true, + "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{job=\"kube-scheduler\",instance=~\"$instance\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{instance}}", + "refId": "A" + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Goroutines", + "tooltip": { + "shared": false, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": true + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": false, + "title": "Dashboard Row", + "titleSize": "h6", + "type": "row" + } + ], + "schemaVersion": 14, + "style": "dark", + "tags": [ + "kubernetes-mixin" + ], + "templating": { + "list": [ + { + "current": { + "text": "Prometheus", + "value": "Prometheus" + }, + "hide": 0, + "label": null, + "name": "datasource", + "options": [ + + ], + "query": "prometheus", + "refresh": 1, + "regex": "", + "type": "datasource" + }, + { + "allValue": null, + "current": { + + }, + "datasource": "$datasource", + "hide": 0, + "includeAll": true, + "label": null, + "multi": false, + "name": "instance", + "options": [ + + ], + "query": "label_values(process_cpu_seconds_total{job=\"kube-scheduler\"}, instance)", + "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": "", + "title": "Kubernetes / Scheduler", + "uid": "2e6b6a3b4bddf1427b3a55aa1311c656", + "version": 0 + } statefulset.json: |- { "__inputs": [ @@ -4555,7 +4728,7 @@ data: "tableColumn": "", "targets": [ { - "expr": "sum(rate(container_cpu_usage_seconds_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod_name=~\"$statefulset.*\"}[3m]))", + "expr": "sum(rate(container_cpu_usage_seconds_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod=~\"$statefulset.*\"}[3m]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "", @@ -4638,7 +4811,7 @@ data: "tableColumn": "", "targets": [ { - "expr": "sum(container_memory_usage_bytes{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod_name=~\"$statefulset.*\"}) / 1024^3", + "expr": "sum(container_memory_usage_bytes{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod=~\"$statefulset.*\"}) / 1024^3", "format": "time_series", "intervalFactor": 2, "legendFormat": "", @@ -4721,7 +4894,7 @@ data: "tableColumn": "", "targets": [ { - "expr": "sum(rate(container_network_transmit_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod_name=~\"$statefulset.*\"}[3m])) + sum(rate(container_network_receive_bytes_total{cluster=\"$cluster\", namespace=\"$namespace\",pod_name=~\"$statefulset.*\"}[3m]))", + "expr": "sum(rate(container_network_transmit_bytes_total{job=\"kubernetes-cadvisor\", cluster=\"$cluster\", namespace=\"$namespace\", pod=~\"$statefulset.*\"}[3m])) + sum(rate(container_network_receive_bytes_total{cluster=\"$cluster\", namespace=\"$namespace\",pod=~\"$statefulset.*\"}[3m]))", "format": "time_series", "intervalFactor": 2, "legendFormat": "", diff --git a/addons/grafana/dashboards-prom.yaml b/addons/grafana/dashboards-prom.yaml new file mode 100644 index 00000000..b1f4c06c --- /dev/null +++ b/addons/grafana/dashboards-prom.yaml @@ -0,0 +1,2160 @@ +apiVersion: v1 +kind: ConfigMap +metadata: + name: grafana-dashboards-prom + namespace: monitoring +data: + prometheus-remote-write.json: |- + { + "annotations": { + "list": [ + + ] + }, + "editable": true, + "gnetId": null, + "graphTooltip": 0, + "hideControls": false, + "links": [ + + ], + "refresh": "10s", + "rows": [ + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 1, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "prometheus_remote_storage_highest_timestamp_in_seconds{cluster=~\"$cluster\", instance=~\"$instance\"} - ignoring(queue) group_right(instance) prometheus_remote_storage_queue_highest_sent_timestamp_seconds{cluster=~\"$cluster\", instance=~\"$instance\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{cluster}}:{{instance}}-{{queue}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Highest Timestamp In vs. Highest Timestamp Sent", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "s", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 2, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "rate(prometheus_remote_storage_highest_timestamp_in_seconds{cluster=~\"$cluster\", instance=~\"$instance\"}[5m]) - ignoring (queue) group_right(instance) rate(prometheus_remote_storage_queue_highest_sent_timestamp_seconds{cluster=~\"$cluster\", instance=~\"$instance\"}[5m])", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{cluster}}:{{instance}}-{{queue}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate[5m]", + "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": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Timestamps", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 3, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "rate(prometheus_remote_storage_samples_in_total{cluster=~\"$cluster\", instance=~\"$instance\"}[5m])- ignoring(queue) group_right(instance) rate(prometheus_remote_storage_succeeded_samples_total{cluster=~\"$cluster\", instance=~\"$instance\"}[5m]) - rate(prometheus_remote_storage_dropped_samples_total{cluster=~\"$cluster\", instance=~\"$instance\"}[5m])", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{cluster}}:{{instance}}-{{queue}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Rate, in vs. succeeded or dropped [5m]", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Samples", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 4, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "prometheus_remote_storage_shards{cluster=~\"$cluster\", instance=~\"$instance\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{cluster}}:{{instance}}-{{queue}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Num. Shards", + "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": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 5, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "prometheus_remote_storage_shard_capacity{cluster=~\"$cluster\", instance=~\"$instance\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{cluster}}:{{instance}}-{{queue}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Capacity", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Shards", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 6, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 3, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "rate(prometheus_remote_storage_dropped_samples_total{cluster=~\"$cluster\", instance=~\"$instance\"}[5m])", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{cluster}}:{{instance}}-{{queue}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Dropped Samples", + "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": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 7, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 3, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "rate(prometheus_remote_storage_failed_samples_total{cluster=~\"$cluster\", instance=~\"$instance\"}[5m])", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{cluster}}:{{instance}}-{{queue}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Failed Samples", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 8, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 3, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "rate(prometheus_remote_storage_retried_samples_total{cluster=~\"$cluster\", instance=~\"$instance\"}[5m])", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{cluster}}:{{instance}}-{{queue}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Retried Samples", + "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": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 9, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 3, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "rate(prometheus_remote_storage_enqueue_retries_total{cluster=~\"$cluster\", instance=~\"$instance\"}[5m])", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{cluster}}:{{instance}}-{{queue}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Enqueue Retries", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Misc Rates.", + "titleSize": "h6" + } + ], + "schemaVersion": 14, + "style": "dark", + "tags": [ + + ], + "templating": { + "list": [ + { + "current": { + "text": "Prometheus", + "value": "Prometheus" + }, + "hide": 0, + "label": null, + "name": "datasource", + "options": [ + + ], + "query": "prometheus", + "refresh": 1, + "regex": "", + "type": "datasource" + }, + { + "allValue": null, + "current": { + "selected": true, + "text": "All", + "value": "$__all" + }, + "datasource": "$datasource", + "hide": 0, + "includeAll": true, + "label": "instance", + "multi": true, + "name": "instance", + "options": [ + + ], + "query": "label_values(prometheus_build_info, instance)", + "refresh": 1, + "regex": "", + "sort": 2, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "current": { + "selected": true, + "text": "All", + "value": "$__all" + }, + "datasource": "$datasource", + "hide": 0, + "includeAll": true, + "label": "cluster", + "multi": true, + "name": "cluster", + "options": [ + + ], + "query": "label_values(kube_pod_container_info{image=~\".*prometheus.*\"}, cluster)", + "refresh": 1, + "regex": "", + "sort": 2, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + } + ] + }, + "time": { + "from": "now-1h", + "to": "now" + }, + "timepicker": { + "refresh_intervals": [ + "5s", + "10s", + "30s", + "1m", + "5m", + "15m", + "30m", + "1h", + "2h", + "1d" + ], + "time_options": [ + "5m", + "15m", + "1h", + "6h", + "12h", + "24h", + "2d", + "7d", + "30d" + ] + }, + "timezone": "utc", + "title": "Prometheus Remote Write", + "uid": "", + "version": 0 + } + prometheus.json: |- + { + "annotations": { + "list": [ + + ] + }, + "editable": true, + "gnetId": null, + "graphTooltip": 0, + "hideControls": false, + "links": [ + + ], + "refresh": "10s", + "rows": [ + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 1, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 12, + "stack": false, + "steppedLine": false, + "styles": [ + { + "alias": "Time", + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "pattern": "Time", + "type": "hidden" + }, + { + "alias": "Count", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "Value #A", + "thresholds": [ + + ], + "type": "hidden", + "unit": "short" + }, + { + "alias": "Uptime", + "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": "short" + }, + { + "alias": "Instance", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "instance", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "Job", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "job", + "thresholds": [ + + ], + "type": "number", + "unit": "short" + }, + { + "alias": "Version", + "colorMode": null, + "colors": [ + + ], + "dateFormat": "YYYY-MM-DD HH:mm:ss", + "decimals": 2, + "link": false, + "linkTooltip": "Drill down", + "linkUrl": "", + "pattern": "version", + "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": "count by (job, instance, version) (prometheus_build_info{job=~\"$job\", instance=~\"$instance\"})", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "A", + "step": 10 + }, + { + "expr": "max by (job, instance) (time() - process_start_time_seconds{job=~\"$job\", instance=~\"$instance\"})", + "format": "table", + "instant": true, + "intervalFactor": 2, + "legendFormat": "", + "refId": "B", + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Prometheus Stats", + "tooltip": { + "shared": true, + "sort": 0, + "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": "Prometheus Stats", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 2, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "sum(rate(prometheus_target_sync_length_seconds_sum{job=~\"$job\",instance=~\"$instance\"}[5m])) by (scrape_job) * 1e3", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{scrape_job}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Target Sync", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "ms", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 3, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "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(prometheus_sd_discovered_targets{job=~\"$job\",instance=~\"$instance\"})", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "Targets", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Targets", + "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": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Discovery", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 1, + "id": 4, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 1, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 4, + "stack": false, + "steppedLine": false, + "targets": [ + { + "expr": "rate(prometheus_target_interval_length_seconds_sum{job=~\"$job\",instance=~\"$instance\"}[5m]) / rate(prometheus_target_interval_length_seconds_count{job=~\"$job\",instance=~\"$instance\"}[5m]) * 1e3", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{interval}} configured", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Average Scrape Interval Duration", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "ms", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 5, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "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": 4, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "sum by (job) (rate(prometheus_target_scrapes_exceeded_sample_limit_total[1m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "exceeded sample limit: {{job}}", + "legendLink": null, + "step": 10 + }, + { + "expr": "sum by (job) (rate(prometheus_target_scrapes_sample_duplicate_timestamp_total[1m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "duplicate timestamp: {{job}}", + "legendLink": null, + "step": 10 + }, + { + "expr": "sum by (job) (rate(prometheus_target_scrapes_sample_out_of_bounds_total[1m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "out of bounds: {{job}}", + "legendLink": null, + "step": 10 + }, + { + "expr": "sum by (job) (rate(prometheus_target_scrapes_sample_out_of_order_total[1m]))", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "out of order: {{job}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Scrape failures", + "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": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 6, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 0, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 4, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "rate(prometheus_tsdb_head_samples_appended_total{job=~\"$job\",instance=~\"$instance\"}[5m])", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{job}} {{instance}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Appended Samples", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Retrieval", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 7, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "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": "prometheus_tsdb_head_series{job=~\"$job\",instance=~\"$instance\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{job}} {{instance}} head series", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Head Series", + "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": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 8, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 0, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "prometheus_tsdb_head_chunks{job=~\"$job\",instance=~\"$instance\"}", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{job}} {{instance}} head chunks", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Head Chunks", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Storage", + "titleSize": "h6" + }, + { + "collapse": false, + "height": "250px", + "panels": [ + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 9, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "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": "rate(prometheus_engine_query_duration_seconds_count{job=~\"$job\",instance=~\"$instance\",slice=\"inner_eval\"}[5m])", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{job}} {{instance}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Query Rate", + "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": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + }, + { + "aliasColors": { + + }, + "bars": false, + "dashLength": 10, + "dashes": false, + "datasource": "$datasource", + "fill": 10, + "id": 10, + "legend": { + "avg": false, + "current": false, + "max": false, + "min": false, + "show": true, + "total": false, + "values": false + }, + "lines": true, + "linewidth": 0, + "links": [ + + ], + "nullPointMode": "null as zero", + "percentage": false, + "pointradius": 5, + "points": false, + "renderer": "flot", + "seriesOverrides": [ + + ], + "spaceLength": 10, + "span": 6, + "stack": true, + "steppedLine": false, + "targets": [ + { + "expr": "max by (slice) (prometheus_engine_query_duration_seconds{quantile=\"0.9\",job=~\"$job\",instance=~\"$instance\"}) * 1e3", + "format": "time_series", + "intervalFactor": 2, + "legendFormat": "{{slice}}", + "legendLink": null, + "step": 10 + } + ], + "thresholds": [ + + ], + "timeFrom": null, + "timeShift": null, + "title": "Stage Duration", + "tooltip": { + "shared": true, + "sort": 0, + "value_type": "individual" + }, + "type": "graph", + "xaxis": { + "buckets": null, + "mode": "time", + "name": null, + "show": true, + "values": [ + + ] + }, + "yaxes": [ + { + "format": "ms", + "label": null, + "logBase": 1, + "max": null, + "min": 0, + "show": true + }, + { + "format": "short", + "label": null, + "logBase": 1, + "max": null, + "min": null, + "show": false + } + ] + } + ], + "repeat": null, + "repeatIteration": null, + "repeatRowId": null, + "showTitle": true, + "title": "Query", + "titleSize": "h6" + } + ], + "schemaVersion": 14, + "style": "dark", + "tags": [ + + ], + "templating": { + "list": [ + { + "current": { + "text": "Prometheus", + "value": "Prometheus" + }, + "hide": 0, + "label": null, + "name": "datasource", + "options": [ + + ], + "query": "prometheus", + "refresh": 1, + "regex": "", + "type": "datasource" + }, + { + "allValue": null, + "current": { + "selected": true, + "text": "All", + "value": "$__all" + }, + "datasource": "$datasource", + "hide": 0, + "includeAll": true, + "label": "job", + "multi": true, + "name": "job", + "options": [ + + ], + "query": "label_values(prometheus_build_info, job)", + "refresh": 1, + "regex": "", + "sort": 2, + "tagValuesQuery": "", + "tags": [ + + ], + "tagsQuery": "", + "type": "query", + "useTags": false + }, + { + "allValue": null, + "current": { + "selected": true, + "text": "All", + "value": "$__all" + }, + "datasource": "$datasource", + "hide": 0, + "includeAll": true, + "label": "instance", + "multi": true, + "name": "instance", + "options": [ + + ], + "query": "label_values(prometheus_build_info, instance)", + "refresh": 1, + "regex": "", + "sort": 2, + "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": "Prometheus", + "uid": "", + "version": 0 + } diff --git a/addons/grafana/deployment.yaml b/addons/grafana/deployment.yaml index dc9f0584..bdc2d891 100644 --- a/addons/grafana/deployment.yaml +++ b/addons/grafana/deployment.yaml @@ -56,8 +56,12 @@ spec: mountPath: /etc/grafana/provisioning/dashboards - name: dashboards-etcd mountPath: /etc/grafana/dashboards/etcd + - name: dashboards-prom + mountPath: /etc/grafana/dashboards/prom - name: dashboards-k8s mountPath: /etc/grafana/dashboards/k8s + - name: dashboards-k8s-nodes + mountPath: /etc/grafana/dashboards/k8s-nodes - name: dashboards-k8s-resources mountPath: /etc/grafana/dashboards/k8s-resources volumes: @@ -73,9 +77,15 @@ spec: - name: dashboards-etcd configMap: name: grafana-dashboards-etcd + - name: dashboards-prom + configMap: + name: grafana-dashboards-prom - name: dashboards-k8s configMap: name: grafana-dashboards-k8s + - name: dashboards-k8s-nodes + configMap: + name: grafana-dashboards-k8s-nodes - name: dashboards-k8s-resources configMap: name: grafana-dashboards-k8s-resources diff --git a/addons/prometheus/rules.yaml b/addons/prometheus/rules.yaml index 2962028a..22e54a91 100644 --- a/addons/prometheus/rules.yaml +++ b/addons/prometheus/rules.yaml @@ -163,32 +163,32 @@ data: "name": "k8s.rules", "rules": [ { - "expr": "sum(rate(container_cpu_usage_seconds_total{job=\"kubernetes-cadvisor\", image!=\"\", container_name!=\"\"}[5m])) by (namespace)\n", + "expr": "sum(rate(container_cpu_usage_seconds_total{job=\"kubernetes-cadvisor\", image!=\"\", container!=\"POD\"}[5m])) by (namespace)\n", "record": "namespace:container_cpu_usage_seconds_total:sum_rate" }, { - "expr": "sum by (namespace, pod_name, container_name) (\n rate(container_cpu_usage_seconds_total{job=\"kubernetes-cadvisor\", image!=\"\", container_name!=\"\"}[5m])\n)\n", - "record": "namespace_pod_name_container_name:container_cpu_usage_seconds_total:sum_rate" + "expr": "sum by (namespace, pod, container) (\n rate(container_cpu_usage_seconds_total{job=\"kubernetes-cadvisor\", image!=\"\", container!=\"POD\"}[5m])\n)\n", + "record": "namespace_pod_container:container_cpu_usage_seconds_total:sum_rate" }, { - "expr": "sum(container_memory_usage_bytes{job=\"kubernetes-cadvisor\", image!=\"\", container_name!=\"\"}) by (namespace)\n", + "expr": "sum(container_memory_usage_bytes{job=\"kubernetes-cadvisor\", image!=\"\", container!=\"POD\"}) by (namespace)\n", "record": "namespace:container_memory_usage_bytes:sum" }, { - "expr": "sum by (namespace, label_name) (\n sum(rate(container_cpu_usage_seconds_total{job=\"kubernetes-cadvisor\", image!=\"\", container_name!=\"\"}[5m])) by (namespace, pod_name)\n * on (namespace, pod_name) group_left(label_name)\n label_replace(kube_pod_labels{job=\"kube-state-metrics\"}, \"pod_name\", \"$1\", \"pod\", \"(.*)\")\n)\n", - "record": "namespace_name:container_cpu_usage_seconds_total:sum_rate" + "expr": "sum by (namespace, label_name) (\n sum(rate(container_cpu_usage_seconds_total{job=\"kubernetes-cadvisor\", image!=\"\", container!=\"POD\"}[5m])) by (namespace, pod)\n * on (namespace, pod)\n group_left(label_name) kube_pod_labels{job=\"kube-state-metrics\"}\n)\n", + "record": "namespace:container_cpu_usage_seconds_total:sum_rate" }, { - "expr": "sum by (namespace, label_name) (\n sum(container_memory_usage_bytes{job=\"kubernetes-cadvisor\",image!=\"\", container_name!=\"\"}) by (pod_name, namespace)\n* on (namespace, pod_name) group_left(label_name)\n label_replace(kube_pod_labels{job=\"kube-state-metrics\"}, \"pod_name\", \"$1\", \"pod\", \"(.*)\")\n)\n", - "record": "namespace_name:container_memory_usage_bytes:sum" + "expr": "sum by (namespace, label_name) (\n sum(container_memory_usage_bytes{job=\"kubernetes-cadvisor\",image!=\"\", container!=\"POD\"}) by (pod, namespace)\n * on (namespace, pod)\n group_left(label_name) kube_pod_labels{job=\"kube-state-metrics\"}\n)\n", + "record": "namespace:container_memory_usage_bytes:sum" }, { - "expr": "sum by (namespace, label_name) (\n sum(kube_pod_container_resource_requests_memory_bytes{job=\"kube-state-metrics\"} * on (endpoint, instance, job, namespace, pod, service) group_left(phase) (kube_pod_status_phase{phase=~\"^(Pending|Running)$\"} == 1)) by (namespace, pod)\n* on (namespace, pod) group_left(label_name)\n label_replace(kube_pod_labels{job=\"kube-state-metrics\"}, \"pod_name\", \"$1\", \"pod\", \"(.*)\")\n)\n", - "record": "namespace_name:kube_pod_container_resource_requests_memory_bytes:sum" + "expr": "sum by (namespace, label_name) (\n sum(kube_pod_container_resource_requests_memory_bytes{job=\"kube-state-metrics\"} * on (endpoint, instance, job, namespace, pod, service) group_left(phase) (kube_pod_status_phase{phase=~\"^(Pending|Running)$\"} == 1)) by (namespace, pod)\n * on (namespace, pod)\n group_left(label_name) kube_pod_labels{job=\"kube-state-metrics\"}\n)\n", + "record": "namespace:kube_pod_container_resource_requests_memory_bytes:sum" }, { - "expr": "sum by (namespace, label_name) (\n sum(kube_pod_container_resource_requests_cpu_cores{job=\"kube-state-metrics\"} * on (endpoint, instance, job, namespace, pod, service) group_left(phase) (kube_pod_status_phase{phase=~\"^(Pending|Running)$\"} == 1)) by (namespace, pod)\n* on (namespace, pod) group_left(label_name)\n label_replace(kube_pod_labels{job=\"kube-state-metrics\"}, \"pod_name\", \"$1\", \"pod\", \"(.*)\")\n)\n", - "record": "namespace_name:kube_pod_container_resource_requests_cpu_cores:sum" + "expr": "sum by (namespace, label_name) (\n sum(kube_pod_container_resource_requests_cpu_cores{job=\"kube-state-metrics\"} * on (endpoint, instance, job, namespace, pod, service) group_left(phase) (kube_pod_status_phase{phase=~\"^(Pending|Running)$\"} == 1)) by (namespace, pod)\n * on (namespace, pod)\n group_left(label_name) kube_pod_labels{job=\"kube-state-metrics\"}\n)\n", + "record": "namespace:kube_pod_container_resource_requests_cpu_cores:sum" }, { "expr": "sum(\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) kube_replicaset_owner{job=\"kube-state-metrics\"},\n \"workload\", \"$1\", \"owner_name\", \"(.*)\"\n )\n) by (namespace, workload, pod)\n", @@ -217,67 +217,67 @@ data: "name": "kube-scheduler.rules", "rules": [ { - "expr": "histogram_quantile(0.99, sum(rate(scheduler_e2e_scheduling_latency_microseconds_bucket{job=\"kube-scheduler\"}[5m])) without(instance, pod)) / 1e+06\n", + "expr": "histogram_quantile(0.99, sum(rate(scheduler_e2e_scheduling_duration_seconds_bucket{job=\"kube-scheduler\"}[5m])) without(instance, pod))\n", "labels": { "quantile": "0.99" }, - "record": "cluster_quantile:scheduler_e2e_scheduling_latency:histogram_quantile" + "record": "cluster_quantile:scheduler_e2e_scheduling_duration_seconds:histogram_quantile" }, { - "expr": "histogram_quantile(0.99, sum(rate(scheduler_scheduling_algorithm_latency_microseconds_bucket{job=\"kube-scheduler\"}[5m])) without(instance, pod)) / 1e+06\n", + "expr": "histogram_quantile(0.99, sum(rate(scheduler_scheduling_algorithm_duration_seconds_bucket{job=\"kube-scheduler\"}[5m])) without(instance, pod))\n", "labels": { "quantile": "0.99" }, - "record": "cluster_quantile:scheduler_scheduling_algorithm_latency:histogram_quantile" + "record": "cluster_quantile:scheduler_scheduling_algorithm_duration_seconds:histogram_quantile" }, { - "expr": "histogram_quantile(0.99, sum(rate(scheduler_binding_latency_microseconds_bucket{job=\"kube-scheduler\"}[5m])) without(instance, pod)) / 1e+06\n", + "expr": "histogram_quantile(0.99, sum(rate(scheduler_binding_duration_seconds_bucket{job=\"kube-scheduler\"}[5m])) without(instance, pod))\n", "labels": { "quantile": "0.99" }, - "record": "cluster_quantile:scheduler_binding_latency:histogram_quantile" + "record": "cluster_quantile:scheduler_binding_duration_seconds:histogram_quantile" }, { - "expr": "histogram_quantile(0.9, sum(rate(scheduler_e2e_scheduling_latency_microseconds_bucket{job=\"kube-scheduler\"}[5m])) without(instance, pod)) / 1e+06\n", + "expr": "histogram_quantile(0.9, sum(rate(scheduler_e2e_scheduling_duration_seconds_bucket{job=\"kube-scheduler\"}[5m])) without(instance, pod))\n", "labels": { "quantile": "0.9" }, - "record": "cluster_quantile:scheduler_e2e_scheduling_latency:histogram_quantile" + "record": "cluster_quantile:scheduler_e2e_scheduling_duration_seconds:histogram_quantile" }, { - "expr": "histogram_quantile(0.9, sum(rate(scheduler_scheduling_algorithm_latency_microseconds_bucket{job=\"kube-scheduler\"}[5m])) without(instance, pod)) / 1e+06\n", + "expr": "histogram_quantile(0.9, sum(rate(scheduler_scheduling_algorithm_duration_seconds_bucket{job=\"kube-scheduler\"}[5m])) without(instance, pod))\n", "labels": { "quantile": "0.9" }, - "record": "cluster_quantile:scheduler_scheduling_algorithm_latency:histogram_quantile" + "record": "cluster_quantile:scheduler_scheduling_algorithm_duration_seconds:histogram_quantile" }, { - "expr": "histogram_quantile(0.9, sum(rate(scheduler_binding_latency_microseconds_bucket{job=\"kube-scheduler\"}[5m])) without(instance, pod)) / 1e+06\n", + "expr": "histogram_quantile(0.9, sum(rate(scheduler_binding_duration_seconds_bucket{job=\"kube-scheduler\"}[5m])) without(instance, pod))\n", "labels": { "quantile": "0.9" }, - "record": "cluster_quantile:scheduler_binding_latency:histogram_quantile" + "record": "cluster_quantile:scheduler_binding_duration_seconds:histogram_quantile" }, { - "expr": "histogram_quantile(0.5, sum(rate(scheduler_e2e_scheduling_latency_microseconds_bucket{job=\"kube-scheduler\"}[5m])) without(instance, pod)) / 1e+06\n", + "expr": "histogram_quantile(0.5, sum(rate(scheduler_e2e_scheduling_duration_seconds_bucket{job=\"kube-scheduler\"}[5m])) without(instance, pod))\n", "labels": { "quantile": "0.5" }, - "record": "cluster_quantile:scheduler_e2e_scheduling_latency:histogram_quantile" + "record": "cluster_quantile:scheduler_e2e_scheduling_duration_seconds:histogram_quantile" }, { - "expr": "histogram_quantile(0.5, sum(rate(scheduler_scheduling_algorithm_latency_microseconds_bucket{job=\"kube-scheduler\"}[5m])) without(instance, pod)) / 1e+06\n", + "expr": "histogram_quantile(0.5, sum(rate(scheduler_scheduling_algorithm_duration_seconds_bucket{job=\"kube-scheduler\"}[5m])) without(instance, pod))\n", "labels": { "quantile": "0.5" }, - "record": "cluster_quantile:scheduler_scheduling_algorithm_latency:histogram_quantile" + "record": "cluster_quantile:scheduler_scheduling_algorithm_duration_seconds:histogram_quantile" }, { - "expr": "histogram_quantile(0.5, sum(rate(scheduler_binding_latency_microseconds_bucket{job=\"kube-scheduler\"}[5m])) without(instance, pod)) / 1e+06\n", + "expr": "histogram_quantile(0.5, sum(rate(scheduler_binding_duration_seconds_bucket{job=\"kube-scheduler\"}[5m])) without(instance, pod))\n", "labels": { "quantile": "0.5" }, - "record": "cluster_quantile:scheduler_binding_latency:histogram_quantile" + "record": "cluster_quantile:scheduler_binding_duration_seconds:histogram_quantile" } ] }, @@ -285,25 +285,25 @@ data: "name": "kube-apiserver.rules", "rules": [ { - "expr": "histogram_quantile(0.99, sum(rate(apiserver_request_latencies_bucket{job=\"apiserver\"}[5m])) without(instance, pod)) / 1e+06\n", + "expr": "histogram_quantile(0.99, sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\"}[5m])) without(instance, pod))\n", "labels": { "quantile": "0.99" }, - "record": "cluster_quantile:apiserver_request_latencies:histogram_quantile" + "record": "cluster_quantile:apiserver_request_duration_seconds:histogram_quantile" }, { - "expr": "histogram_quantile(0.9, sum(rate(apiserver_request_latencies_bucket{job=\"apiserver\"}[5m])) without(instance, pod)) / 1e+06\n", + "expr": "histogram_quantile(0.9, sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\"}[5m])) without(instance, pod))\n", "labels": { "quantile": "0.9" }, - "record": "cluster_quantile:apiserver_request_latencies:histogram_quantile" + "record": "cluster_quantile:apiserver_request_duration_seconds:histogram_quantile" }, { - "expr": "histogram_quantile(0.5, sum(rate(apiserver_request_latencies_bucket{job=\"apiserver\"}[5m])) without(instance, pod)) / 1e+06\n", + "expr": "histogram_quantile(0.5, sum(rate(apiserver_request_duration_seconds_bucket{job=\"apiserver\"}[5m])) without(instance, pod))\n", "labels": { "quantile": "0.5" }, - "record": "cluster_quantile:apiserver_request_latencies:histogram_quantile" + "record": "cluster_quantile:apiserver_request_duration_seconds:histogram_quantile" } ] }, @@ -403,11 +403,11 @@ data: "record": "node:node_disk_saturation:avg_irate" }, { - "expr": "max by (namespace, pod, device) ((node_filesystem_size_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\"}\n- node_filesystem_avail_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\"})\n/ node_filesystem_size_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\"})\n", + "expr": "max by (instance, namespace, pod, device) ((node_filesystem_size_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\"}\n- node_filesystem_avail_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\"})\n/ node_filesystem_size_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\"})\n", "record": "node:node_filesystem_usage:" }, { - "expr": "max by (namespace, pod, device) (node_filesystem_avail_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\"} / node_filesystem_size_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\"})\n", + "expr": "max by (instance, namespace, pod, device) (node_filesystem_avail_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\"} / node_filesystem_size_bytes{fstype=~\"ext[234]|btrfs|xfs|zfs\"})\n", "record": "node:node_filesystem_avail:" }, { @@ -510,7 +510,7 @@ data: "message": "Pod {{ $labels.namespace }}/{{ $labels.pod }} has been in a non-ready state for longer than an hour.", "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubepodnotready" }, - "expr": "sum by (namespace, pod) (kube_pod_status_phase{job=\"kube-state-metrics\", phase=~\"Pending|Unknown\"}) > 0\n", + "expr": "sum by (namespace, pod) (kube_pod_status_phase{job=\"kube-state-metrics\", phase=~\"Failed|Pending|Unknown\"}) > 0\n", "for": "1h", "labels": { "severity": "critical" @@ -659,7 +659,7 @@ data: "message": "Cluster has overcommitted CPU resource requests for Pods and cannot tolerate node failure.", "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubecpuovercommit" }, - "expr": "sum(namespace_name:kube_pod_container_resource_requests_cpu_cores:sum)\n /\nsum(node:node_num_cpu:sum)\n >\n(count(node:node_num_cpu:sum)-1) / count(node:node_num_cpu:sum)\n", + "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", "labels": { "severity": "warning" @@ -671,7 +671,7 @@ data: "message": "Cluster has overcommitted memory resource requests for Pods and cannot tolerate node failure.", "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubememovercommit" }, - "expr": "sum(namespace_name:kube_pod_container_resource_requests_memory_bytes:sum)\n /\nsum(node_memory_MemTotal_bytes)\n >\n(count(node:node_num_cpu:sum)-1)\n /\ncount(node:node_num_cpu:sum)\n", + "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", "labels": { "severity": "warning" @@ -683,7 +683,7 @@ data: "message": "Cluster has overcommitted CPU resource requests for Namespaces.", "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubecpuovercommit" }, - "expr": "sum(kube_resourcequota{job=\"kube-state-metrics\", type=\"hard\", resource=\"cpu\"})\n /\nsum(node:node_num_cpu:sum)\n > 1.5\n", + "expr": "sum(kube_resourcequota{job=\"kube-state-metrics\", type=\"hard\", resource=\"cpu\"})\n /\nsum(kube_node_status_allocatable_cpu_cores)\n > 1.5\n", "for": "5m", "labels": { "severity": "warning" @@ -695,7 +695,7 @@ data: "message": "Cluster has overcommitted memory resource requests for Namespaces.", "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubememovercommit" }, - "expr": "sum(kube_resourcequota{job=\"kube-state-metrics\", type=\"hard\", resource=\"memory\"})\n /\nsum(node_memory_MemTotal_bytes{job=\"node-exporter\"})\n > 1.5\n", + "expr": "sum(kube_resourcequota{job=\"kube-state-metrics\", type=\"hard\", resource=\"memory\"})\n /\nsum(kube_node_status_allocatable_memory_bytes{job=\"node-exporter\"})\n > 1.5\n", "for": "5m", "labels": { "severity": "warning" @@ -716,10 +716,10 @@ data: { "alert": "CPUThrottlingHigh", "annotations": { - "message": "{{ printf \"%0.0f\" $value }}% throttling of CPU in namespace {{ $labels.namespace }} for container {{ $labels.container_name }} in pod {{ $labels.pod_name }}.", + "message": "{{ printf \"%0.0f\" $value }}% throttling of CPU in namespace {{ $labels.namespace }} for container {{ $labels.container }} in pod {{ $labels.pod }}.", "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-cputhrottlinghigh" }, - "expr": "100 * sum(increase(container_cpu_cfs_throttled_periods_total{container_name!=\"\", }[5m])) by (container_name, pod_name, namespace)\n /\nsum(increase(container_cpu_cfs_periods_total{}[5m])) by (container_name, pod_name, namespace)\n > 100 \n", + "expr": "100 * sum(increase(container_cpu_cfs_throttled_periods_total{container!=\"\", }[5m])) by (container, pod, namespace)\n /\nsum(increase(container_cpu_cfs_periods_total{}[5m])) by (container, pod, namespace)\n > 100 \n", "for": "15m", "labels": { "severity": "warning" @@ -837,7 +837,7 @@ data: "message": "The API server has a 99th percentile latency of {{ $value }} seconds for {{ $labels.verb }} {{ $labels.resource }}.", "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeapilatencyhigh" }, - "expr": "cluster_quantile:apiserver_request_latencies:histogram_quantile{job=\"apiserver\",quantile=\"0.99\",subresource!=\"log\",verb!~\"^(?:LIST|WATCH|WATCHLIST|PROXY|CONNECT)$\"} > 1\n", + "expr": "cluster_quantile:apiserver_request_duration_seconds:histogram_quantile{job=\"apiserver\",quantile=\"0.99\",subresource!=\"log\",verb!~\"^(?:LIST|WATCH|WATCHLIST|PROXY|CONNECT)$\"} > 1\n", "for": "10m", "labels": { "severity": "warning" @@ -849,7 +849,7 @@ data: "message": "The API server has a 99th percentile latency of {{ $value }} seconds for {{ $labels.verb }} {{ $labels.resource }}.", "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeapilatencyhigh" }, - "expr": "cluster_quantile:apiserver_request_latencies:histogram_quantile{job=\"apiserver\",quantile=\"0.99\",subresource!=\"log\",verb!~\"^(?:LIST|WATCH|WATCHLIST|PROXY|CONNECT)$\"} > 4\n", + "expr": "cluster_quantile:apiserver_request_duration_seconds:histogram_quantile{job=\"apiserver\",quantile=\"0.99\",subresource!=\"log\",verb!~\"^(?:LIST|WATCH|WATCHLIST|PROXY|CONNECT)$\"} > 4\n", "for": "10m", "labels": { "severity": "critical" @@ -861,7 +861,7 @@ data: "message": "API server is returning errors for {{ $value }}% of requests.", "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeapierrorshigh" }, - "expr": "sum(rate(apiserver_request_count{job=\"apiserver\",code=~\"^(?:5..)$\"}[5m]))\n /\nsum(rate(apiserver_request_count{job=\"apiserver\"}[5m])) * 100 > 3\n", + "expr": "sum(rate(apiserver_request_total{job=\"apiserver\",code=~\"^(?:5..)$\"}[5m]))\n /\nsum(rate(apiserver_request_total{job=\"apiserver\"}[5m])) * 100 > 3\n", "for": "10m", "labels": { "severity": "critical" @@ -873,7 +873,7 @@ data: "message": "API server is returning errors for {{ $value }}% of requests.", "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeapierrorshigh" }, - "expr": "sum(rate(apiserver_request_count{job=\"apiserver\",code=~\"^(?:5..)$\"}[5m]))\n /\nsum(rate(apiserver_request_count{job=\"apiserver\"}[5m])) * 100 > 1\n", + "expr": "sum(rate(apiserver_request_total{job=\"apiserver\",code=~\"^(?:5..)$\"}[5m]))\n /\nsum(rate(apiserver_request_total{job=\"apiserver\"}[5m])) * 100 > 1\n", "for": "10m", "labels": { "severity": "warning" @@ -885,7 +885,7 @@ data: "message": "API server is returning errors for {{ $value }}% of requests for {{ $labels.verb }} {{ $labels.resource }} {{ $labels.subresource }}.", "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeapierrorshigh" }, - "expr": "sum(rate(apiserver_request_count{job=\"apiserver\",code=~\"^(?:5..)$\"}[5m])) by (resource,subresource,verb)\n /\nsum(rate(apiserver_request_count{job=\"apiserver\"}[5m])) by (resource,subresource,verb) * 100 > 10\n", + "expr": "sum(rate(apiserver_request_total{job=\"apiserver\",code=~\"^(?:5..)$\"}[5m])) by (resource,subresource,verb)\n /\nsum(rate(apiserver_request_total{job=\"apiserver\"}[5m])) by (resource,subresource,verb) * 100 > 10\n", "for": "10m", "labels": { "severity": "critical" @@ -897,7 +897,7 @@ data: "message": "API server is returning errors for {{ $value }}% of requests for {{ $labels.verb }} {{ $labels.resource }} {{ $labels.subresource }}.", "runbook_url": "https://github.com/kubernetes-monitoring/kubernetes-mixin/tree/master/runbook.md#alert-name-kubeapierrorshigh" }, - "expr": "sum(rate(apiserver_request_count{job=\"apiserver\",code=~\"^(?:5..)$\"}[5m])) by (resource,subresource,verb)\n /\nsum(rate(apiserver_request_count{job=\"apiserver\"}[5m])) by (resource,subresource,verb) * 100 > 5\n", + "expr": "sum(rate(apiserver_request_total{job=\"apiserver\",code=~\"^(?:5..)$\"}[5m])) by (resource,subresource,verb)\n /\nsum(rate(apiserver_request_total{job=\"apiserver\"}[5m])) by (resource,subresource,verb) * 100 > 5\n", "for": "10m", "labels": { "severity": "warning" @@ -1046,131 +1046,6 @@ data: } ] }, - { - "name": "prometheus.rules", - "rules": [ - { - "alert": "PrometheusConfigReloadFailed", - "annotations": { - "description": "Reloading Prometheus' configuration has failed for {{$labels.namespace}}/{{$labels.pod}}", - "summary": "Reloading Prometheus' configuration failed" - }, - "expr": "prometheus_config_last_reload_successful{job=\"prometheus\"} == 0\n", - "for": "10m", - "labels": { - "severity": "warning" - } - }, - { - "alert": "PrometheusNotificationQueueRunningFull", - "annotations": { - "description": "Prometheus' alert notification queue is running full for {{$labels.namespace}}/{{ $labels.pod}}", - "summary": "Prometheus' alert notification queue is running full" - }, - "expr": "predict_linear(prometheus_notifications_queue_length{job=\"prometheus\"}[5m], 60 * 30) > prometheus_notifications_queue_capacity{job=\"prometheus\"}\n", - "for": "10m", - "labels": { - "severity": "warning" - } - }, - { - "alert": "PrometheusErrorSendingAlerts", - "annotations": { - "description": "Errors while sending alerts from Prometheus {{$labels.namespace}}/{{ $labels.pod}} to Alertmanager {{$labels.Alertmanager}}", - "summary": "Errors while sending alert from Prometheus" - }, - "expr": "rate(prometheus_notifications_errors_total{job=\"prometheus\"}[5m]) / rate(prometheus_notifications_sent_total{job=\"prometheus\"}[5m]) > 0.01\n", - "for": "10m", - "labels": { - "severity": "warning" - } - }, - { - "alert": "PrometheusErrorSendingAlerts", - "annotations": { - "description": "Errors while sending alerts from Prometheus {{$labels.namespace}}/{{ $labels.pod}} to Alertmanager {{$labels.Alertmanager}}", - "summary": "Errors while sending alerts from Prometheus" - }, - "expr": "rate(prometheus_notifications_errors_total{job=\"prometheus\"}[5m]) / rate(prometheus_notifications_sent_total{job=\"prometheus\"}[5m]) > 0.03\n", - "for": "10m", - "labels": { - "severity": "critical" - } - }, - { - "alert": "PrometheusNotConnectedToAlertmanagers", - "annotations": { - "description": "Prometheus {{ $labels.namespace }}/{{ $labels.pod}} is not connected to any Alertmanagers", - "summary": "Prometheus is not connected to any Alertmanagers" - }, - "expr": "prometheus_notifications_alertmanagers_discovered{job=\"prometheus\"} < 1\n", - "for": "10m", - "labels": { - "severity": "warning" - } - }, - { - "alert": "PrometheusTSDBReloadsFailing", - "annotations": { - "description": "{{$labels.job}} at {{$labels.instance}} had {{$value | humanize}} reload failures over the last four hours.", - "summary": "Prometheus has issues reloading data blocks from disk" - }, - "expr": "increase(prometheus_tsdb_reloads_failures_total{job=\"prometheus\"}[2h]) > 0\n", - "for": "12h", - "labels": { - "severity": "warning" - } - }, - { - "alert": "PrometheusTSDBCompactionsFailing", - "annotations": { - "description": "{{$labels.job}} at {{$labels.instance}} had {{$value | humanize}} compaction failures over the last four hours.", - "summary": "Prometheus has issues compacting sample blocks" - }, - "expr": "increase(prometheus_tsdb_compactions_failed_total{job=\"prometheus\"}[2h]) > 0\n", - "for": "12h", - "labels": { - "severity": "warning" - } - }, - { - "alert": "PrometheusTSDBWALCorruptions", - "annotations": { - "description": "{{$labels.job}} at {{$labels.instance}} has a corrupted write-ahead log (WAL).", - "summary": "Prometheus write-ahead log is corrupted" - }, - "expr": "prometheus_tsdb_wal_corruptions_total{job=\"prometheus\"} > 0\n", - "for": "4h", - "labels": { - "severity": "warning" - } - }, - { - "alert": "PrometheusNotIngestingSamples", - "annotations": { - "description": "Prometheus {{ $labels.namespace }}/{{ $labels.pod}} isn't ingesting samples.", - "summary": "Prometheus isn't ingesting samples" - }, - "expr": "rate(prometheus_tsdb_head_samples_appended_total{job=\"prometheus\"}[5m]) <= 0\n", - "for": "10m", - "labels": { - "severity": "warning" - } - }, - { - "alert": "PrometheusTargetScrapesDuplicate", - "annotations": { - "description": "{{$labels.namespace}}/{{$labels.pod}} has many samples rejected due to duplicate timestamps but different values", - "summary": "Prometheus has many samples rejected" - }, - "expr": "increase(prometheus_target_scrapes_sample_duplicate_timestamp_total{job=\"prometheus\"}[5m]) > 0\n", - "for": "10m", - "labels": { - "severity": "warning" - } - } - ] - }, { "name": "general.rules", "rules": [ @@ -1189,3 +1064,193 @@ data: } ] } + prom.yaml: |- + { + "groups": [ + { + "name": "prometheus", + "rules": [ + { + "alert": "PrometheusBadConfig", + "annotations": { + "description": "Prometheus {{$labels.instance}} has failed to reload its configuration.", + "summary": "Failed Prometheus configuration reload." + }, + "expr": "# Without max_over_time, failed scrapes could create false negatives, see\n# https://www.robustperception.io/alerting-on-gauges-in-prometheus-2-0 for details.\nmax_over_time(prometheus_config_last_reload_successful{job=\"prometheus\"}[5m]) == 0\n", + "for": "10m", + "labels": { + "severity": "critical" + } + }, + { + "alert": "PrometheusNotificationQueueRunningFull", + "annotations": { + "description": "Alert notification queue of Prometheus {{$labels.instance}} is running full.", + "summary": "Prometheus alert notification queue predicted to run full in less than 30m." + }, + "expr": "# Without min_over_time, failed scrapes could create false negatives, see\n# https://www.robustperception.io/alerting-on-gauges-in-prometheus-2-0 for details.\n(\n predict_linear(prometheus_notifications_queue_length{job=\"prometheus\"}[5m], 60 * 30)\n>\n min_over_time(prometheus_notifications_queue_capacity{job=\"prometheus\"}[5m])\n)\n", + "for": "15m", + "labels": { + "severity": "warning" + } + }, + { + "alert": "PrometheusErrorSendingAlertsToSomeAlertmanagers", + "annotations": { + "description": "{{ printf \"%.1f\" $value }}% errors while sending alerts from Prometheus {{$labels.instance}} to Alertmanager {{$labels.alertmanager}}.", + "summary": "Prometheus has encountered more than 1% errors sending alerts to a specific Alertmanager." + }, + "expr": "(\n rate(prometheus_notifications_errors_total{job=\"prometheus\"}[5m])\n/\n rate(prometheus_notifications_sent_total{job=\"prometheus\"}[5m])\n)\n* 100\n> 1\n", + "for": "15m", + "labels": { + "severity": "warning" + } + }, + { + "alert": "PrometheusErrorSendingAlertsToAnyAlertmanager", + "annotations": { + "description": "{{ printf \"%.1f\" $value }}% minimum errors while sending alerts from Prometheus {{$labels.instance}} to any Alertmanager.", + "summary": "Prometheus encounters more than 3% errors sending alerts to any Alertmanager." + }, + "expr": "min without(alertmanager) (\n rate(prometheus_notifications_errors_total{job=\"prometheus\"}[5m])\n/\n rate(prometheus_notifications_sent_total{job=\"prometheus\"}[5m])\n)\n* 100\n> 3\n", + "for": "15m", + "labels": { + "severity": "critical" + } + }, + { + "alert": "PrometheusNotConnectedToAlertmanagers", + "annotations": { + "description": "Prometheus {{$labels.instance}} is not connected to any Alertmanagers.", + "summary": "Prometheus is not connected to any Alertmanagers." + }, + "expr": "# Without max_over_time, failed scrapes could create false negatives, see\n# https://www.robustperception.io/alerting-on-gauges-in-prometheus-2-0 for details.\nmax_over_time(prometheus_notifications_alertmanagers_discovered{job=\"prometheus\"}[5m]) < 1\n", + "for": "10m", + "labels": { + "severity": "warning" + } + }, + { + "alert": "PrometheusTSDBReloadsFailing", + "annotations": { + "description": "Prometheus {{$labels.instance}} has detected {{$value | humanize}} reload failures over the last 3h.", + "summary": "Prometheus has issues reloading blocks from disk." + }, + "expr": "increase(prometheus_tsdb_reloads_failures_total{job=\"prometheus\"}[3h]) > 0\n", + "for": "4h", + "labels": { + "severity": "warning" + } + }, + { + "alert": "PrometheusTSDBCompactionsFailing", + "annotations": { + "description": "Prometheus {{$labels.instance}} has detected {{$value | humanize}} compaction failures over the last 3h.", + "summary": "Prometheus has issues compacting blocks." + }, + "expr": "increase(prometheus_tsdb_compactions_failed_total{job=\"prometheus\"}[3h]) > 0\n", + "for": "4h", + "labels": { + "severity": "warning" + } + }, + { + "alert": "PrometheusTSDBWALCorruptions", + "annotations": { + "description": "Prometheus {{$labels.instance}} has detected {{$value | humanize}} corruptions of the write-ahead log (WAL) over the last 3h.", + "summary": "Prometheus is detecting WAL corruptions." + }, + "expr": "increase(tsdb_wal_corruptions_total{job=\"prometheus\"}[3h]) > 0\n", + "for": "4h", + "labels": { + "severity": "warning" + } + }, + { + "alert": "PrometheusNotIngestingSamples", + "annotations": { + "description": "Prometheus {{$labels.instance}} is not ingesting samples.", + "summary": "Prometheus is not ingesting samples." + }, + "expr": "rate(prometheus_tsdb_head_samples_appended_total{job=\"prometheus\"}[5m]) <= 0\n", + "for": "10m", + "labels": { + "severity": "warning" + } + }, + { + "alert": "PrometheusDuplicateTimestamps", + "annotations": { + "description": "Prometheus {{$labels.instance}} is dropping {{$value | humanize}} samples/s with different values but duplicated timestamp.", + "summary": "Prometheus is dropping samples with duplicate timestamps." + }, + "expr": "rate(prometheus_target_scrapes_sample_duplicate_timestamp_total{job=\"prometheus\"}[5m]) > 0\n", + "for": "10m", + "labels": { + "severity": "warning" + } + }, + { + "alert": "PrometheusOutOfOrderTimestamps", + "annotations": { + "description": "Prometheus {{$labels.instance}} is dropping {{$value | humanize}} samples/s with timestamps arriving out of order.", + "summary": "Prometheus drops samples with out-of-order timestamps." + }, + "expr": "rate(prometheus_target_scrapes_sample_out_of_order_total{job=\"prometheus\"}[5m]) > 0\n", + "for": "10m", + "labels": { + "severity": "warning" + } + }, + { + "alert": "PrometheusRemoteStorageFailures", + "annotations": { + "description": "Prometheus {{$labels.instance}} failed to send {{ printf \"%.1f\" $value }}% of the samples to queue {{$labels.queue}}.", + "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", + "for": "15m", + "labels": { + "severity": "critical" + } + }, + { + "alert": "PrometheusRemoteWriteBehind", + "annotations": { + "description": "Prometheus {{$labels.instance}} remote write is {{ printf \"%.1f\" $value }}s behind for queue {{$labels.queue}}.", + "summary": "Prometheus remote write is behind." + }, + "expr": "# Without max_over_time, failed scrapes could create false negatives, see\n# https://www.robustperception.io/alerting-on-gauges-in-prometheus-2-0 for details.\n(\n max_over_time(prometheus_remote_storage_highest_timestamp_in_seconds{job=\"prometheus\"}[5m])\n- on(job, instance) group_right\n max_over_time(prometheus_remote_storage_queue_highest_sent_timestamp_seconds{job=\"prometheus\"}[5m])\n)\n> 120\n", + "for": "15m", + "labels": { + "severity": "critical" + } + }, + { + "alert": "PrometheusRuleFailures", + "annotations": { + "description": "Prometheus {{$labels.instance}} has failed to evaluate {{ printf \"%.0f\" $value }} rules in the last 5m.", + "summary": "Prometheus is failing rule evaluations." + }, + "expr": "increase(prometheus_rule_evaluation_failures_total{job=\"prometheus\"}[5m]) > 0\n", + "for": "15m", + "labels": { + "severity": "critical" + } + }, + { + "alert": "PrometheusMissingRuleEvaluations", + "annotations": { + "description": "Prometheus {{$labels.instance}} has missed {{ printf \"%.0f\" $value }} rule group evaluations in the last 5m.", + "summary": "Prometheus is missing rule evaluations due to slow rule group evaluation." + }, + "expr": "increase(prometheus_rule_group_iterations_missed_total{job=\"prometheus\"}[5m]) > 0\n", + "for": "15m", + "labels": { + "severity": "warning" + } + } + ] + } + ] + }