K3S Installation

Cluster Nodes

The cluster consists of four Raspberry Pi 4 (8 GB RAM, 32 GB SD card), each with a fixed role:

Node	Role
cube01	Worker
cube02	Worker
cube03	Worker
cube04	Control plane

Network Configuration (optional)

By default, nodes are identified through the router’s DNS. If that’s not available, static IPs can be assigned manually.

On each node, edit /etc/network/interfaces:

auto eth0
iface eth0 inet static
    address <node-ip>
    netmask 255.255.255.0
    gateway <gateway-ip>
    broadcast <broadcast-ip>
    hostname <node-name>

Then edit /etc/hosts so each node knows the others:

127.0.0.1       localhost
::1             localhost ip6-localhost ip6-loopback
ff02::1         ip6-allnodes
ff02::2         ip6-allrouters

127.0.1.1       <node-name>
<IP-cube01>     cube01 cube01.local
<IP-cube02>     cube02 cube02.local
<IP-cube03>     cube03 cube03.local
<IP-cube04>     cube04 cube04.local

System Preparation

These steps apply to every node.

Enable cgroups

Append to the end of /boot/firmware/cmdline.txt (on a single line), then reboot:

cgroup_enable=cpuset cgroup_enable=memory cgroup_memory=1

Hostname

sudo nano /etc/hostname
# Set the node name: cube01, cube02, etc.

Passwordless SSH Access

Copy the control plane’s (cube04) public SSH key into each worker’s ~/.ssh/authorized_keys to allow password-free connections.

Install iptables

sudo apt -y install iptables

Disable swap (control plane only)

Kubernetes requires swap to be disabled:

sudo swapoff -a
sudo nano /etc/dphys-swapfile
# Set CONF_SWAPSIZE=0

K3S Installation

Control plane (cube04)

curl -sfL https://get.k3s.io | INSTALL_K3S_EXEC="server" sh -s - \
  --write-kubeconfig-mode 644 \
  --node-taint CriticalAddonsOnly=true:NoExecute \
  --bind-address <IP-cube04>

The --node-taint CriticalAddonsOnly=true:NoExecute option prevents application pods from being scheduled on the control plane.

Then retrieve the worker registration token:

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

Workers (cube01, cube02, cube03)

On each worker, install the k3s agent pointing to the control plane:

curl -sfL https://get.k3s.io | INSTALL_K3S_EXEC="agent" \
  K3S_URL=https://<IP-cube04>:6443 \
  K3S_TOKEN="<K3S_TOKEN>" sh -

Post-Installation

Node Labels

From the control plane, label each worker so workloads are correctly targeted:

kubectl label nodes cube01 kubernetes.io/role=worker node-type=worker
kubectl label nodes cube02 kubernetes.io/role=worker node-type=worker
kubectl label nodes cube03 kubernetes.io/role=worker node-type=worker

KUBECONFIG Variable

Add to /etc/environment on the control plane so Helm and other tools find the configuration:

KUBECONFIG=/etc/rancher/k3s/k3s.yaml

Install Helm

curl -fsSL -o get_helm.sh https://raw.githubusercontent.com/helm/helm/main/scripts/get-helm-3
chmod 700 get_helm.sh
./get_helm.sh

Initial Cluster State

At this point, all four nodes are registered and system pods are running. kubectl get nodes should show all four nodes with Ready status.