Migrate GCP, DO, Azure to static pod control plane

* Run a kube-apiserver, kube-scheduler, and kube-controller-manager
static pod on each controller node. Previously, kube-apiserver was
self-hosted as a DaemonSet across controllers and kube-scheduler
and kube-controller-manager were a Deployment (with 2 or
controller_count many replicas).
* Remove bootkube bootstrap and pivot to self-hosted
* Remove pod-checkpointer manifests (no longer needed)

azure/container-linux/kubernetes/README.md

@@ -11,7 +11,7 @@ Typhoon distributes upstream Kubernetes, architectural conventions, and cluster
 
 ## Features <a href="https://www.cncf.io/certification/software-conformance/"><img align="right" src="https://storage.googleapis.com/poseidon/certified-kubernetes.png"></a>
 
-* Kubernetes v1.15.3 (upstream, via [kubernetes-incubator/bootkube](https://github.com/kubernetes-incubator/bootkube))
+* Kubernetes v1.15.3 (upstream)
 * Single or multi-master, [Calico](https://www.projectcalico.org/) or [flannel](https://github.com/coreos/flannel) networking
 * On-cluster etcd with TLS, [RBAC](https://kubernetes.io/docs/admin/authorization/rbac/)-enabled, [network policy](https://kubernetes.io/docs/concepts/services-networking/network-policies/)
 * Advanced features like [worker pools](https://typhoon.psdn.io/advanced/worker-pools/), [low-priority](https://typhoon.psdn.io/cl/azure/#low-priority) workers, and [snippets](https://typhoon.psdn.io/advanced/customization/#container-linux) customization

azure/container-linux/kubernetes/bootkube.tf

@@ -1,6 +1,6 @@
 # Self-hosted Kubernetes assets (kubeconfig, manifests)
 module "bootkube" {
-  source = "git::https://github.com/poseidon/terraform-render-bootkube.git?ref=98cc19f80f2c4c3ddc63fc7aea6320e74bec561a"
+  source = "git::https://github.com/poseidon/terraform-render-bootkube.git?ref=6e59af71138bc5f784453873074de16e7ee150eb"
 
   cluster_name = var.cluster_name
   api_servers  = [format("%s.%s", var.cluster_name, var.dns_zone)]

azure/container-linux/kubernetes/cl/controller.yaml.tmpl

@@ -96,17 +96,28 @@ systemd:
         RestartSec=10
         [Install]
         WantedBy=multi-user.target
-    - name: bootkube.service
+    - name: bootstrap.service
       contents: |
         [Unit]
-        Description=Bootstrap a Kubernetes cluster
-        ConditionPathExists=!/opt/bootkube/init_bootkube.done
+        Description=Kubernetes control plane
+        ConditionPathExists=!/opt/bootstrap/bootstrap.done
         [Service]
         Type=oneshot
         RemainAfterExit=true
-        WorkingDirectory=/opt/bootkube
-        ExecStart=/opt/bootkube/bootkube-start
-        ExecStartPost=/bin/touch /opt/bootkube/init_bootkube.done
+        WorkingDirectory=/opt/bootstrap
+        ExecStartPre=-/usr/bin/bash -c 'set -x && [ -n "$(ls /opt/bootstrap/assets/manifests-*/* 2>/dev/null)" ] && mv /opt/bootstrap/assets/manifests-*/* /opt/bootstrap/assets/manifests && rm -rf /opt/bootstrap/assets/manifests-*'
+        ExecStart=/usr/bin/rkt run \
+            --trust-keys-from-https \
+            --volume assets,kind=host,source=/opt/bootstrap/assets \
+            --mount volume=assets,target=/assets \
+            --volume script,kind=host,source=/opt/bootstrap/apply \
+            --mount volume=script,target=/apply \
+            --insecure-options=image \
+            docker://k8s.gcr.io/hyperkube:v1.15.3 \
+            --net=host \
+            --dns=host \
+            --exec=/apply
+        ExecStartPost=/bin/touch /opt/bootstrap/bootstrap.done
         [Install]
         WantedBy=multi-user.target
 storage:
@@ -124,36 +135,26 @@ storage:
         inline: |
           KUBELET_IMAGE_URL=docker://k8s.gcr.io/hyperkube
           KUBELET_IMAGE_TAG=v1.15.3
+    - path: /opt/bootstrap/apply
+      filesystem: root
+      mode: 0544
+      contents:
+        inline: |
+          #!/bin/bash -e
+          export KUBECONFIG=/assets/auth/kubeconfig
+          until kubectl version; do
+            echo "Waiting for static pod control plane"
+            sleep 5
+          done
+          until kubectl apply -f /assets/manifests -R; do
+             echo "Retry applying manifests"
+             sleep 5
+          done
     - path: /etc/sysctl.d/max-user-watches.conf
       filesystem: root
       contents:
         inline: |
           fs.inotify.max_user_watches=16184
-    - path: /opt/bootkube/bootkube-start
-      filesystem: root
-      mode: 0544
-      user:
-        id: 500
-      group:
-        id: 500
-      contents:
-        inline: |
-          #!/bin/bash
-          # Wrapper for bootkube start
-          set -e
-          # Move experimental manifests
-          [ -n "$(ls /opt/bootkube/assets/manifests-*/* 2>/dev/null)" ] && mv /opt/bootkube/assets/manifests-*/* /opt/bootkube/assets/manifests && rm -rf /opt/bootkube/assets/manifests-*
-          exec /usr/bin/rkt run \
-            --trust-keys-from-https \
-            --volume assets,kind=host,source=/opt/bootkube/assets \
-            --mount volume=assets,target=/assets \
-            --volume bootstrap,kind=host,source=/etc/kubernetes \
-            --mount volume=bootstrap,target=/etc/kubernetes \
-            $${RKT_OPTS} \
-            quay.io/coreos/bootkube:v0.14.0 \
-            --net=host \
-            --dns=host \
-            --exec=/bootkube -- start --asset-dir=/assets "$@"
 passwd:
   users:
     - name: core

azure/container-linux/kubernetes/security.tf

@@ -53,6 +53,22 @@ resource "azurerm_network_security_rule" "controller-etcd-metrics" {
   destination_address_prefix  = azurerm_subnet.controller.address_prefix
 }
 
+# Allow Prometheus to scrape kube-scheduler and kube-controller-manager metrics
+resource "azurerm_network_security_rule" "controller-kube-metrics" {
+  resource_group_name = azurerm_resource_group.cluster.name
+
+  name                        = "allow-kube-metrics"
+  network_security_group_name = azurerm_network_security_group.controller.name
+  priority                    = "2011"
+  access                      = "Allow"
+  direction                   = "Inbound"
+  protocol                    = "Tcp"
+  source_port_range           = "*"
+  destination_port_range      = "10251-10252"
+  source_address_prefix       = azurerm_subnet.worker.address_prefix
+  destination_address_prefix  = azurerm_subnet.controller.address_prefix
+}
+
 resource "azurerm_network_security_rule" "controller-apiserver" {
   resource_group_name = azurerm_resource_group.cluster.name
 

azure/container-linux/kubernetes/ssh.tf

@@ -1,12 +1,15 @@
-# Secure copy etcd TLS assets to controllers.
+# Secure copy assets to controllers.
 resource "null_resource" "copy-controller-secrets" {
   count = var.controller_count
 
-  depends_on = [azurerm_virtual_machine.controllers]
+  depends_on = [
+    module.bootkube,
+    azurerm_virtual_machine.controllers
+  ]
 
   connection {
     type    = "ssh"
-    host    = element(azurerm_public_ip.controllers.*.ip_address, count.index)
+    host    = azurerm_public_ip.controllers.*.ip_address[count.index]
     user    = "core"
     timeout = "15m"
   }
@@ -45,6 +48,11 @@ resource "null_resource" "copy-controller-secrets" {
     content     = module.bootkube.etcd_peer_key
     destination = "$HOME/etcd-peer.key"
   }
+  
+  provisioner "file" {
+    source      = var.asset_dir
+    destination = "$HOME/assets"
+  }
 
   provisioner "remote-exec" {
     inline = [
@@ -58,18 +66,21 @@ resource "null_resource" "copy-controller-secrets" {
       "sudo mv etcd-peer.key /etc/ssl/etcd/etcd/peer.key",
       "sudo chown -R etcd:etcd /etc/ssl/etcd",
       "sudo chmod -R 500 /etc/ssl/etcd",
+      "sudo mv $HOME/assets /opt/bootstrap/assets",
+      "sudo mkdir -p /etc/kubernetes/bootstrap-secrets",
+      "sudo cp -r /opt/bootstrap/assets/tls/* /etc/kubernetes/bootstrap-secrets/",
+      "sudo cp /opt/bootstrap/assets/auth/kubeconfig /etc/kubernetes/bootstrap-secrets/",
+      "sudo cp -r /opt/bootstrap/assets/static-manifests/* /etc/kubernetes/manifests/",
     ]
   }
 }
 
-# Secure copy bootkube assets to ONE controller and start bootkube to perform
-# one-time self-hosted cluster bootstrapping.
-resource "null_resource" "bootkube-start" {
+# Connect to a controller to perform one-time cluster bootstrap.
+resource "null_resource" "bootstrap" {
   depends_on = [
-    module.bootkube,
+    null_resource.copy-controller-secrets,
     module.workers,
     azurerm_dns_a_record.apiserver,
-    null_resource.copy-controller-secrets,
   ]
 
   connection {
@@ -79,15 +90,9 @@ resource "null_resource" "bootkube-start" {
     timeout = "15m"
   }
 
-  provisioner "file" {
-    source      = var.asset_dir
-    destination = "$HOME/assets"
-  }
-
   provisioner "remote-exec" {
     inline = [
-      "sudo mv $HOME/assets /opt/bootkube",
-      "sudo systemctl start bootkube",
+      "sudo systemctl start bootstrap",
     ]
   }
 }