k3s-alpine

Tools to set up k3s on a Raspberry Pi running Alpine Linux.

Prerequisites

• Raspberry Pi 4
• SD card and means of attaching it to a computer
• Linux computer that will do the flashing of the cards

Usage

Settings

Download urls and target files are set in settings.yaml.

K3s first server

Before installing any agents, you need to install and init a server. It will use the k3s config in config/server-init.yaml.

1. Prepare the SD card

   • Plug in the SD card and check its name with lsblk

     NOTE: next steps will wipe the drive, so make sure you find the correct one

   • Format it and make two drives - one boot partition (FAT32) and one for k3s data (e.g. ext4)

     sudo parted <DRIVE NAME> rm 1  rm 2  mkpart primary FAT32 1M 1024M  mkpart primary FAT32 1024M 100%  set 1 boot on  print
     mkfs.fat <DRIVE NAME>1
     mkfs.ext4 <DRIVE NAME>2

     Note that this is already done withing the boot make target, but please check the code before running it, as it might inadvertently delete data.

     (see parted docs)

2. Build and install the server

   • Prepare a config file, see: config/server-init.yaml

     NOTE: you need to set k3s.config.cluster-init: true in the config for the first control plane, but only for the first one.

   • Mount the boot partition: mkdir -p /mnt/sd && mount <DRIVE NAME>1 /mnt/sd

   • Build and install: make install path=/mnt/sd config=config/server-init.yaml

   • Unmount the partition: umount /mnt/sd

3. Insert the SD card into the Raspberry Pi and power it up. After a minute, the init script should install k3s and start it up.

Additional nodes

4. Once the server is up, ssh onto the Pi and get the agent token that will be used to add more nodes:

   sudo cat /var/lib/rancher/k3s/server/agent-token

   Set the IP and token into the k3s-configs/agent.yaml file.

5. Prepare another SD card following steps 1-3, except in step 2, where the make command needs the agent config file: make install path=/mnt/sd config=config/agent.yaml

   NOTE: If you're adding another control plan, set k3s.exec: server in the config without k3s.config.cluster-init: true and with IP and token of the existing server.

6. The agent should join the cluster after booting.

Kubectl

Docs: https://docs.k3s.io/cluster-access

• From control plane:

  ssh user@control-plane-ip
  sudo k3s kubectl get nodes

• From outside the cluster:

  ssh user@control-plane-ip
  sudo cat /etc/rancher/k3s/k3s.yaml
  

  Copy the contents of the file to $HOME/.kube/config or as one of the contexts into $HOME/.kube/contexts/two-tauers.yaml.

  Or use this handy command to copy from the control plane called sauron and rename the cluster and user with yq:

  ssh sauron sudo cat /etc/rancher/k3s/k3s.yaml | yq e '.clusters.0.cluster.server="https://192.168.0.100:6443" | .clusters.0.name="two-tauers" | .contexts.0.context.cluster="two-tauers" | .contexts.0.context.user="tt" | .contexts.0.name="two-tauers" | .users.0.name="tt" | .current-context="two-tauers"' > ~/.kube/contexts/two-tauers.yaml