This Ansible playbook turns your fleet of Raspberry Pi into Kubernetes cattle with k3s, the streamlined version of Kubernetes by @rancher that runs silky smooth on the ARM processor powering your Pi.

Manually prepare your Raspberries

• Burn Raspian Stretch Lite on an SD card with EtcherBalena, or something alike
• Create an empty ssh file in the root of the SD card (volume is called boot)

Secure your Raspberries

Optional but advisable: use SSH key auth and disable password login. The default ssh credentials for a Raspberry Pi are username pi and password raspberry.

• Copy your pubkey to the Pi with ssh-copy-id pi@kube1.example.com
• In /etc/ssh/sshd_config on the Pi, set PasswordAuthentication to no

Configure ansible setup

• Copy ansible/hosts.template to ansible/hosts for your configuration
• Choose one of the Raspberries to lead / orchestrate your cluster; we'll call this the server
• In ansible/hosts, fill out the ip or hostname for the leading Raspberry under [k3s-server]
• Fill out the ip's or hostnames for the rest of your cattle under [k3s-agents]

I personally prefer using the Pi's hardware mac address to assign a hostname and ip address within the LAN by DHCP, but you could also set a static ip address on the Pi.

Provision the nodes

When you're all set up and configured, run the Ansible playbook:

$ make

This will:

• Install the k3s binary on the 'server' Pi (the leading node)
• Install a k3s-server service on the server and start it
• Fetch the node token from the server
• Install and start a k3s-agent service on the agents, joining the cluster
• Enable autostart on boot for k3s on all nodes

See if it worked

Log into your server node and run:

$ sudo k3s kubectl get node -o wide

You should see all of your nodes broadcasting a Ready status.

Happy cattle herding!