To setup your Rancher Cluster, run the following:

./setup_cluster.sh

The script will output the K3S_TOKEN to a file called .k3s_token which is going to be used to bake the image.

There is a terraform module named packer_module that will run a docker image and bake an image. main.tf uses it and that is what is CircleCI uses as well as what is going to be run locally.

To bake an image, create a module like the one below. We are going to use the one defined in terraform/example as an example.

module "build_image" {
  source = "./packer_module"

  wifi_name     = var.wifi_name
  wifi_password = var.wifi_password
  ssh_key_path  = var.ssh_key_path
  master_ip     = var.master_ip
  k3s_token     = var.k3s_token
}

Here are what the terraform variables do:

• $WIFI_NAME is the name of the wifi network that the Raspberry Pis are going to connect to.
• $WIFI_PASSWORD is the password of the wifi network if there is one.
• $SSH_KEY_PATH is the path of your public key to add to the authorized_keys of the image.
• $K3S_TOKEN is the rancher provided token when creating a cluster. The default location is /var/lib/rancher/k3s/server/node-token, or if the setup_cluster script was used, .k3s_token.
• $MASTER is the address to connect to the master. It needs the master port, which is normally 6443.

flasher has also been provided to image an sd card. Run the command below (Assuming /dev/sdb/ is your sd card) to image the card.

sudo ./flasher -image=./output-arm-image/image -device=/dev/sdb

Assuming kubectl is installed on the machine that ran the ./setup_cluster.sh command, setting the following environment variable to gain access to the cluster.

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

Using the image above, any Raspberry Pis that startup using that image will automatically join your cluster.