K8S Kubeadm Cluster with Terraform

This repository creates a kubeadm cluster with the Hetzner cloud.

The preconfigured components are etcd, containerd and calico.

Prerequisites

• Hetzner Cloud account
• Hetzner Cloud access token.
• Terraform CLI

Usage

Export your Hetzner cloud access token as an environment variable, select the staging workspace, and start deploying.

export TF_VAR_hcloud_token=SECRET

terraform init
terraform workspace new staging
terraform apply

Then, access your cluster via SSH and grab the kubeconfig file:

ssh root@${$(terraform output controller_ip)//\"/} \
  -o StrictHostKeyChecking=no \
  -i .ssh-${$(terraform workspace show)//\"/}/id_rsa.key \
  -c "cat /root/.kube/config"

Customization

Kubernetes Version

The default value is v1.26.4, change it in bin/01_install.sh:

KUBERNETES_VERSION=1.26.4

Cluster Nodes

The variables.tf file contains the configuration which Hetzner server type to use and the number and name for the controller and worker nodes. It also distinguishes into two workspaces with different number of nodes:

• staging: 1 controller (cx21), 2 worker nodes (cpx21)
• production: 1 controller (cx31), 4 worker nodes (cpx31)

Edit the variables.tf to change this.

Switch the workspace with terraform workspace select staging or terraform workspace select production, and use terraform refresh to get the current controller IP address.

Known Limitations

• Currently, only 1 controller node can be provisioned
• The default user on the node is root