merged k3s-install into master

.gitignore

@@ -4,4 +4,9 @@ crash.log
 kubeconfig.yaml
 kubeconfig.yaml-e
 terraform.tfvars
+<<<<<<< HEAD
 plans.yaml
+=======
+plans.yaml
+traefik_config.yaml
+>>>>>>> k3s-install

README.md

@@ -128,20 +128,26 @@ By default, we have 3 control planes configured and 2 agents, with automatic upg
 
 **But if you want to remain HA, it's important to keep a number of control planes nodes of at least 3 (2 to maintain quorum when 1 goes down for automated upgrades and reboot for instance), see [Rancher's doc on HA](https://rancher.com/docs/k3s/latest/en/installation/ha-embedded/).**
 
-Otherwise, it's important to turn off automatic upgrades (see below) and reboots for the control-plane nodes (2 or less), and do the maintenance yourself.
+Otherwise, it's important to turn off automatic upgrades (see below) for the control-plane nodes (2 or less), and do the maintenance yourself.
 
 ## Automatic upgrade
 
-By default, MicroOS and its embedded k3s instance get upgraded automatically on each node, and reboot safely via [Kured](https://github.com/weaveworks/kured) installed in the cluster.
+By default, MicroOS gets upgraded automatically on each node, and reboot safely via [Kured](https://github.com/weaveworks/kured) installed in the cluster.
 
-_About [Kured](https://github.com/weaveworks/kured), it does not have a latest tag present for its image, but it's pretty compatible, so you can just manually update the tag from once every year for instance._
+As for k3s it is also automatically upgrades thanks to Rancher's <https://github.com/rancher/system-upgrade-controller>. By default it follows the k3s `stable` channel, but you can also change to `latest` one if needed, or specify a target version to upgrade to via the upgrade plan. You can copy and modify the one in the templates for that! More on the subject in [k3s upgrades basic](https://rancher.com/docs/k3s/latest/en/upgrades/basic/).
 
-_Last but not least, if you wish to turn off automatic upgrade on a specific node, you need to ssh into it and issue the following command:_
+_If you wish to turn off automatic MicroOS upgrades on a specific node, you need to ssh into it and issue the following command:_
 
 ```sh
 systemctl --now disable transactional-update.timer
 ```
 
+_To turn off k3s upgrades, you can either set the `k3s_upgrade=true` label in the node you want, or set it to `false`. To just remove it, apply:_
+
+```sh
+kubectl -n system-upgrade label node <node-name> k3s_upgrade-
+```
+
 ## Takedown
 
 If you want to takedown the cluster, you can proceed as follows:

agents.tf

@@ -31,17 +31,21 @@ resource "hcloud_server" "agents" {
     destination = "/root/config.ign"
   }
 
+  # Combustion script file to install k3s-selinux
+  provisioner "file" {
+    content     = local.combustion_script
+    destination = "/root/script"
+  }
+
   # Install MicroOS
   provisioner "remote-exec" {
-    inline = local.MicroOS_install_commands
+    inline = local.microOS_install_commands
   }
 
-  # Issue a reboot command
+  # Issue a reboot command and wait for the node to reboot
   provisioner "local-exec" {
-    command = "ssh ${local.ssh_args} root@${self.ipv4_address} '(sleep 2; reboot)&'; sleep 3"
+    command = "ssh ${local.ssh_args} root@${self.ipv4_address} '(sleep 2; reboot)&'; sleep 5"
   }
-
-  # Wait for MicroOS to reboot and be ready
   provisioner "local-exec" {
     command = <<-EOT
       until ssh ${local.ssh_args} -o ConnectTimeout=2 root@${self.ipv4_address} true 2> /dev/null
@@ -52,50 +56,42 @@ resource "hcloud_server" "agents" {
     EOT
   }
 
-
-  # Generating and uploading the agent.conf file
-  provisioner "file" {
-    content = templatefile("${path.module}/templates/agent.conf.tpl", {
-      server = "https://${local.first_control_plane_network_ip}:6443"
-      token  = random_password.k3s_token.result
-    })
-    destination = "/etc/rancher/k3s/agent.conf"
-  }
-
   # Generating k3s agent config file
   provisioner "file" {
     content = yamlencode({
       node-name     = self.name
+      server        = "https://${local.first_control_plane_network_ip}:6443"
+      token         = random_password.k3s_token.result
       kubelet-arg   = "cloud-provider=external"
       flannel-iface = "eth1"
-      node-ip       = cidrhost(hcloud_network_subnet.k3s.ip_range, 257 + count.index)
+      node-ip       = cidrhost(hcloud_network_subnet.k3s.ip_range, 513 + count.index)
+      node-label    = var.automatically_upgrade_k3s ? ["k3s_upgrade=true"] : []
     })
-    destination = "/etc/rancher/k3s/config.yaml"
+    destination = "/tmp/config.yaml"
   }
 
-  # Run the agent
+  # Install k3s agent
+  provisioner "remote-exec" {
+    inline = local.install_k3s_agent
+  }
+
+  # Upon reboot verify that k3s agent starts correctly
   provisioner "remote-exec" {
     inline = [
-      # set the hostname in a persistent fashion
-      "hostnamectl set-hostname ${self.name}",
-      # first we disable automatic reboot (after transactional updates), and configure the reboot method as kured
-      "rebootmgrctl set-strategy off && echo 'REBOOT_METHOD=kured' > /etc/transactional-update.conf",
-      # then we start k3s agent and join the cluster
-      "systemctl enable k3s-agent",
       <<-EOT
-        until systemctl status k3s-agent > /dev/null
+      timeout 120 bash <<EOF
-        do
+        until systemctl status k3s-agent > /dev/null; do
-          systemctl start k3s-agent
+          echo "Waiting for the k3s agent to start..."
-          echo "Starting k3s-agent and joining the cluster..."
           sleep 2
         done
+      EOF
       EOT
     ]
   }
 
   network {
     network_id = hcloud_network.k3s.id
-    ip         = cidrhost(hcloud_network_subnet.k3s.ip_range, 257 + count.index)
+    ip         = cidrhost(hcloud_network_subnet.k3s.ip_range, 513 + count.index)
   }
 
   depends_on = [

locals.tf

@@ -1,5 +1,5 @@
 locals {
-  first_control_plane_network_ip = cidrhost(hcloud_network_subnet.k3s.ip_range, 2)
+  first_control_plane_network_ip = cidrhost(hcloud_network_subnet.k3s.ip_range, 257)
   hcloud_image_name              = "ubuntu-20.04"
 
   ssh_public_key = trimspace(file(var.public_key))
@@ -18,11 +18,12 @@ locals {
   csi_version   = var.hetzner_csi_version != null ? var.hetzner_csi_version : data.github_release.hetzner_csi.release_tag
   kured_version = data.github_release.kured.release_tag
 
-  MicroOS_install_commands = [
+  microOS_install_commands = [
     "set -ex",
+    "apt-get update",
     "apt-get install -y aria2",
-    "aria2c --follow-metalink=mem https://download.opensuse.org/tumbleweed/appliances/openSUSE-MicroOS.x86_64-k3s-kvm-and-xen.qcow2.meta4",
+    "aria2c --follow-metalink=mem https://download.opensuse.org/tumbleweed/appliances/openSUSE-MicroOS.x86_64-kvm-and-xen.qcow2.meta4",
-    "qemu-img convert -p -f qcow2 -O host_device $(ls -a | grep -ie '^opensuse.*microos.*k3s.*qcow2$') /dev/sda",
+    "qemu-img convert -p -f qcow2 -O host_device $(ls -a | grep -ie '^opensuse.*microos.*qcow2$') /dev/sda",
     "sgdisk -e /dev/sda",
     "parted -s /dev/sda resizepart 4 99%",
     "parted -s /dev/sda mkpart primary ext2 99% 100%",
@@ -32,6 +33,31 @@ locals {
     "mount /dev/sda5 /mnt",
     "mkdir /mnt/ignition",
     "cp /root/config.ign /mnt/ignition/config.ign",
+    "mkdir /mnt/combustion",
+    "cp /root/script /mnt/combustion/script",
     "umount /mnt"
   ]
+
+  combustion_script = <<EOF
+#!/bin/bash
+# combustion: network
+rpm --import https://rpm.rancher.io/public.key
+zypper refresh
+zypper --gpg-auto-import-keys install -y https://rpm.rancher.io/k3s/stable/common/microos/noarch/k3s-selinux-0.4-1.sle.noarch.rpm
+udevadm settle
+    EOF
+
+  common_commands_install_k3s = [
+    "set -ex",
+    # first we disable automatic reboot (after transactional updates), and configure the reboot method as kured
+    "rebootmgrctl set-strategy off && echo 'REBOOT_METHOD=kured' > /etc/transactional-update.conf",
+    # prepare the k3s config directory
+    "mkdir -p /etc/rancher/k3s",
+    # move the config file into place
+    "mv /tmp/config.yaml /etc/rancher/k3s/config.yaml"
+  ]
+
+  install_k3s_server = concat(local.common_commands_install_k3s, ["curl -sfL https://get.k3s.io | INSTALL_K3S_SKIP_SELINUX_RPM=true INSTALL_K3S_SKIP_START=true INSTALL_K3S_EXEC=server sh -"])
+
+  install_k3s_agent = concat(local.common_commands_install_k3s, ["curl -sfL https://get.k3s.io | INSTALL_K3S_SKIP_SELINUX_RPM=true INSTALL_K3S_EXEC=agent sh -"])
 }

master.tf

@@ -29,17 +29,21 @@ resource "hcloud_server" "first_control_plane" {
     destination = "/root/config.ign"
   }
 
+  # Combustion script file to install k3s-selinux
+  provisioner "file" {
+    content     = local.combustion_script
+    destination = "/root/script"
+  }
+
   # Install MicroOS
   provisioner "remote-exec" {
-    inline = local.MicroOS_install_commands
+    inline = local.microOS_install_commands
   }
 
-  # Issue a reboot command
+  # Issue a reboot command and wait for the node to reboot
   provisioner "local-exec" {
-    command = "ssh ${local.ssh_args} root@${self.ipv4_address} '(sleep 2; reboot)&'; sleep 3"
+    command = "ssh ${local.ssh_args} root@${self.ipv4_address} '(sleep 2; reboot)&'; sleep 5"
   }
-
-  # Wait for MicroOS to reboot and be ready
   provisioner "local-exec" {
     command = <<-EOT
       until ssh ${local.ssh_args} -o ConnectTimeout=2 root@${self.ipv4_address} true 2> /dev/null
@@ -63,36 +67,36 @@ resource "hcloud_server" "first_control_plane" {
       advertise-address        = local.first_control_plane_network_ip
       token                    = random_password.k3s_token.result
       node-taint               = var.allow_scheduling_on_control_plane ? [] : ["node-role.kubernetes.io/master:NoSchedule"]
+      node-label               = var.automatically_upgrade_k3s ? ["k3s_upgrade=true"] : []
     })
-    destination = "/etc/rancher/k3s/config.yaml"
+    destination = "/tmp/config.yaml"
   }
 
-  # Run the first control plane
+  # Install k3s server
+  provisioner "remote-exec" {
+    inline = local.install_k3s_server
+  }
+
+  # Upon reboot verify that the k3s server is starts, and wait for k3s to be ready to receive commands
   provisioner "remote-exec" {
     inline = [
-      # set the hostname in a persistent fashion
+      "systemctl start k3s",
-      "hostnamectl set-hostname ${self.name}",
+      # prepare the post_install directory
-      # first we disable automatic reboot (after transactional updates), and configure the reboot method as kured
-      "rebootmgrctl set-strategy off && echo 'REBOOT_METHOD=kured' > /etc/transactional-update.conf",
-      # prepare a directory for our post-installation kustomizations
       "mkdir -p /tmp/post_install",
-      # then we initiate the cluster
+      # wait for k3s to become ready
-      "systemctl enable k3s-server",
-      # wait for k3s to get ready
       <<-EOT
       timeout 120 bash <<EOF
-        until systemctl status k3s-server > /dev/null; do
+        until systemctl status k3s > /dev/null; do
-          systemctl start k3s-server
+          echo "Waiting for the k3s server to start..."
-          echo "Initiating the cluster..."
+          sleep 2
-          sleep 1
         done
         until [ -e /etc/rancher/k3s/k3s.yaml ]; do
           echo "Waiting for kubectl config..."
-          sleep 1
+          sleep 2
         done
         until [[ "\$(kubectl get --raw='/readyz' 2> /dev/null)" == "ok" ]]; do
           echo "Waiting for the cluster to become ready..."
-          sleep 1
+          sleep 2
         done
       EOF
       EOT
@@ -108,11 +112,13 @@ resource "hcloud_server" "first_control_plane" {
         "https://github.com/hetznercloud/hcloud-cloud-controller-manager/releases/download/${local.ccm_version}/ccm-networks.yaml",
         "https://raw.githubusercontent.com/hetznercloud/csi-driver/${local.csi_version}/deploy/kubernetes/hcloud-csi.yml",
         "https://github.com/weaveworks/kured/releases/download/${local.kured_version}/kured-${local.kured_version}-dockerhub.yaml",
-        "./traefik.yaml"
+        "https://raw.githubusercontent.com/rancher/system-upgrade-controller/master/manifests/system-upgrade-controller.yaml",
+        "traefik.yaml",
       ]
       patchesStrategicMerge = [
         file("${path.module}/patches/kured.yaml"),
-        file("${path.module}/patches/ccm.yaml")
+        file("${path.module}/patches/ccm.yaml"),
+        file("${path.module}/patches/system-upgrade-controller.yaml")
       ]
     })
     destination = "/tmp/post_install/kustomization.yaml"
@@ -132,9 +138,20 @@ resource "hcloud_server" "first_control_plane" {
     destination = "/tmp/post_install/traefik.yaml"
   }
 
+  # Upload the system upgrade controller plans config
+  provisioner "file" {
+    content = templatefile(
+      "${path.module}/templates/plans.yaml.tpl",
+      {
+        channel = var.k3s_upgrade_channel
+    })
+    destination = "/tmp/post_install/plans.yaml"
+  }
+
   # Deploy secrets, logging is automatically disabled due to sensitive variables
   provisioner "remote-exec" {
     inline = [
+      "set -ex",
       "kubectl -n kube-system create secret generic hcloud --from-literal=token=${var.hcloud_token} --from-literal=network=${hcloud_network.k3s.name}",
       "kubectl -n kube-system create secret generic hcloud-csi --from-literal=token=${var.hcloud_token}",
     ]
@@ -143,6 +160,7 @@ resource "hcloud_server" "first_control_plane" {
   # Deploy our post-installation kustomization
   provisioner "remote-exec" {
     inline = [
+      "set -ex",
       # This ugly hack is here, because terraform serializes the
       # embedded yaml files with "- |2", when there is more than
       # one yamldocument in the embedded file. Kustomize does not understand
@@ -153,6 +171,9 @@ resource "hcloud_server" "first_control_plane" {
       # manifests themselves
       "sed -i 's/^- |[0-9]\\+$/- |/g' /tmp/post_install/kustomization.yaml",
       "kubectl apply -k /tmp/post_install",
+      "echo 'Waiting for the system-upgrade-controller deployment to become available...'",
+      "kubectl -n system-upgrade wait --for=condition=available --timeout=120s deployment/system-upgrade-controller",
+      "kubectl -n system-upgrade apply -f /tmp/post_install/plans.yaml"
     ]
   }
 

patches/system-upgrade-controller.yaml

@@ -0,0 +1,18 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: system-upgrade-controller
+  namespace: system-upgrade
+spec:
+  template:
+    spec:
+      containers:
+        - name: system-upgrade-controller
+          volumeMounts:
+            - name: ca-certificates
+              mountPath: /var/lib/ca-certificates
+      volumes:
+        - name: ca-certificates
+          hostPath:
+            path: /var/lib/ca-certificates
+            type: Directory

servers.tf

@@ -30,17 +30,21 @@ resource "hcloud_server" "control_planes" {
     destination = "/root/config.ign"
   }
 
+  # Combustion script file to install k3s-selinux
+  provisioner "file" {
+    content     = local.combustion_script
+    destination = "/root/script"
+  }
+
   # Install MicroOS
   provisioner "remote-exec" {
-    inline = local.MicroOS_install_commands
+    inline = local.microOS_install_commands
   }
 
-  # Issue a reboot command
+  # Issue a reboot command and wait for the node to reboot
   provisioner "local-exec" {
-    command = "ssh ${local.ssh_args} root@${self.ipv4_address} '(sleep 2; reboot)&'; sleep 3"
+    command = "ssh ${local.ssh_args} root@${self.ipv4_address} '(sleep 2; reboot)&'; sleep 5"
   }
-
-  # Wait for MicroOS to reboot and be ready
   provisioner "local-exec" {
     command = <<-EOT
       until ssh ${local.ssh_args} -o ConnectTimeout=2 root@${self.ipv4_address} true 2> /dev/null
@@ -56,43 +60,44 @@ resource "hcloud_server" "control_planes" {
     content = yamlencode({
       node-name                = self.name
       server                   = "https://${local.first_control_plane_network_ip}:6443"
+      token                    = random_password.k3s_token.result
       cluster-init             = true
       disable-cloud-controller = true
       disable                  = "servicelb, local-storage"
       flannel-iface            = "eth1"
       kubelet-arg              = "cloud-provider=external"
-      node-ip                  = cidrhost(hcloud_network_subnet.k3s.ip_range, 3 + count.index)
+      node-ip                  = cidrhost(hcloud_network_subnet.k3s.ip_range, 258 + count.index)
-      advertise-address        = cidrhost(hcloud_network_subnet.k3s.ip_range, 3 + count.index)
+      advertise-address        = cidrhost(hcloud_network_subnet.k3s.ip_range, 258 + count.index)
-      tls-san                  = cidrhost(hcloud_network_subnet.k3s.ip_range, 3 + count.index)
+      tls-san                  = cidrhost(hcloud_network_subnet.k3s.ip_range, 258 + count.index)
-      token                    = random_password.k3s_token.result
       node-taint               = var.allow_scheduling_on_control_plane ? [] : ["node-role.kubernetes.io/master:NoSchedule"]
+      node-label               = var.automatically_upgrade_k3s ? ["k3s_upgrade=true"] : []
     })
-    destination = "/etc/rancher/k3s/config.yaml"
+    destination = "/tmp/config.yaml"
   }
 
-  # Run an other control plane server
+  # Install k3s server
+  provisioner "remote-exec" {
+    inline = local.install_k3s_server
+  }
+
+  # Upon reboot verify that the k3s server starts correctly
   provisioner "remote-exec" {
     inline = [
-      # set the hostname in a persistent fashion
+      "systemctl start k3s",
-      "hostnamectl set-hostname ${self.name}",
-      # first we disable automatic reboot (after transactional updates), and configure the reboot method as kured
-      "rebootmgrctl set-strategy off && echo 'REBOOT_METHOD=kured' > /etc/transactional-update.conf",
-      # then then we start k3s in server mode and join the cluster
-      "systemctl enable k3s-server",
       <<-EOT
-        until systemctl status k3s-server > /dev/null
+      timeout 120 bash <<EOF
-        do
+        until systemctl status k3s > /dev/null; do
-          systemctl start k3s-server
+          echo "Waiting for the k3s server to start..."
-          echo "Waiting on other 'learning' control planes, patience is the mother of all virtues..."
           sleep 2
         done
+      EOF
       EOT
     ]
   }
 
   network {
     network_id = hcloud_network.k3s.id
-    ip         = cidrhost(hcloud_network_subnet.k3s.ip_range, 3 + count.index)
+    ip         = cidrhost(hcloud_network_subnet.k3s.ip_range, 258 + count.index)
   }
 
   depends_on = [

templates/agent.conf.tpl

@@ -1,3 +0,0 @@
-SERVER_URL="${server}"
-NODE_TOKEN="${token}"
-AGENT_OPTS=""

templates/plans.yaml.tpl

@@ -0,0 +1,50 @@
+# Doc: https://rancher.com/docs/k3s/latest/en/upgrades/automated/
+# agent plan
+apiVersion: upgrade.cattle.io/v1
+kind: Plan
+metadata:
+  name: k3s-agent
+  namespace: system-upgrade
+  labels:
+    k3s_upgrade: agent
+spec:
+  concurrency: 1
+  channel: https://update.k3s.io/v1-release/channels/${channel}
+  nodeSelector:
+    matchExpressions:
+      - {key: k3s_upgrade, operator: Exists}
+      - {key: k3s_upgrade, operator: NotIn, values: ["disabled", "false"]}
+      - {key: node-role.kubernetes.io/master, operator: NotIn, values: ["true"]}
+  serviceAccountName: system-upgrade
+  prepare:
+    image: rancher/k3s-upgrade
+    args: ["prepare", "k3s-server"]
+  drain:
+    force: true
+    skipWaitForDeleteTimeout: 60
+  upgrade:
+    image: rancher/k3s-upgrade
+---
+# server plan
+apiVersion: upgrade.cattle.io/v1
+kind: Plan
+metadata:
+  name: k3s-server
+  namespace: system-upgrade
+  labels:
+    k3s_upgrade: server
+spec:
+  concurrency: 1
+  channel: https://update.k3s.io/v1-release/channels/${channel}
+  nodeSelector:
+    matchExpressions:
+      - {key: k3s_upgrade, operator: Exists}
+      - {key: k3s_upgrade, operator: NotIn, values: ["disabled", "false"]}
+      - {key: node-role.kubernetes.io/master, operator: In, values: ["true"]}
+  tolerations:
+    - {key: node-role.kubernetes.io/master, effect: NoSchedule, operator: Exists}
+    - {key: CriticalAddonsOnly, effect: NoExecute, operator: Exists}
+  serviceAccountName: system-upgrade
+  cordon: true
+  upgrade:
+    image: rancher/k3s-upgrade

templates/traefik_config.yaml.tpl

@@ -10,9 +10,9 @@ spec:
       type: LoadBalancer
       annotations:
         "load-balancer.hetzner.cloud/name": "traefik"
-        # make hetzners load-balancer connect to our nodes via our private k3s-net.
+        # make hetzners load-balancer connect to our nodes via our private k3s
         "load-balancer.hetzner.cloud/use-private-ip": "true"
-        # keep hetzner-ccm from exposing our private ingress ip, which in general isn't routeable from the public internet.
+        # keep hetzner-ccm from exposing our private ingress ip, which in general isn't routeable from the public internet
         "load-balancer.hetzner.cloud/disable-private-ingress": "true"
         # disable ipv6 by default, because external-dns doesn't support AAAA for hcloud yet https://github.com/kubernetes-sigs/external-dns/issues/2044
         "load-balancer.hetzner.cloud/ipv6-disabled": "${lb_disable_ipv6}"

terraform.tfvars.example

@@ -29,3 +29,10 @@ agents_num                = 2
 
 # If you want to allow non-control-plane workloads to run on the control-plane nodes set "true" below. The default is "false".
 # allow_scheduling_on_control_plane = true
+
+# If you want to disable automatic upgrade of k3s (stable channel), you can set this to false, default is "true".
+# automatically_upgrade_k3s = false
+
+# If you would like to specify the k3s upgrade channel from the get go, you can do so, the default is "stable".
+# For a list of available channels, see https://rancher.com/docs/k3s/latest/en/upgrades/basic/ and https://update.k3s.io/v1-release/channels
+# k3s_upgrade_channel = "latest"

variables.tf

@@ -84,3 +84,15 @@ variable "allow_scheduling_on_control_plane" {
   default     = false
   description = "Whether to allow non-control-plane workloads to run on the control-plane nodes"
 }
+
+variable "k3s_upgrade_channel" {
+  type        = string
+  default     = "stable"
+  description = "Allows you to specify the k3s upgrade channel"
+}
+
+variable "automatically_upgrade_k3s" {
+  type        = bool
+  default     = true
+  description = "Whether to automatically upgrade k3s based on the selected channel"
+}