RaspberryPi.md

Raspberry Pi

You can run this just fine on one or more raspberry pis.

The Pi 5 works really well :)

Architecture

The pi architecture is arm.

I would suggest going with the arm64 OS and installing it on an SSD (USB SSD is fine).

From the docs - If deploying K3s with embedded etcd on a Raspberry Pi, it is recommended that you use an external SSD. etcd is write intensive, and SD cards cannot handle the IO load.

Setup

You must enable cgroups.

From the docs:

Append cgroup_memory=1 cgroup_enable=memory to /boot/cmdline.txt after building the OS image before booting it.

You may also need to install vxlan support.

apt install linux-modules-extra-raspi

If its not found - check does modprobe vxlan return any errors - if not - you already have it.

Multiplatform

Intel images from dockerhub will not run - it requires arm.

More and more images are being built multiplatform.

Here's a snippet of a github action that builds multiplatform:

- name: Set up QEMU
  uses: docker/setup-qemu-action@v3

- name: Set up Docker Buildx
  uses: docker/setup-buildx-action@v3

- name: Login to DockerHub Registry
  uses: docker/login-action@v3
  with:
    username: ${{ secrets.DOCKERHUB_USERNAME }}
    password: ${{ secrets.DOCKERHUB_TOKEN }}

- name: Build and push
  uses: docker/build-push-action@v5
  with:
    context: .
    platforms: linux/amd64,linux/arm64
    push: true
  tags: "${{ secrets.DOCKERHUB_USERNAME }}/${{ github.event.repository.name }}:${{ github.sha }},${{ secrets.DOCKERHUB_USERNAME }}/${{ github.event.repository.name }}:latest"

Affinity

K3S sets a bunch of labels on each node by default.

Amongst them are some about the OS - e.g. on a Pi 5 it shows:

• kubernetes.io/arch=arm64
• kubernetes.io/os=linux

So - if you have a mixed cluster - some arm64 and some amd64 you can set node affinity:

For example - to force a deployment to an intel node:

affinity:
  nodeAffinity:
    requiredDuringSchedulingIgnoredDuringExecution:
      nodeSelectorTerms:
        - matchExpressions:
            - key: arch
              operator: In
              values:
                - amd64

This blog post was easy to understand.