Bootstrapping a 6-Node HA kubeadm Cluster on WSL2 with Terraform

I wanted a homelab cluster that actually looked like production — multiple control-plane nodes, a load balancer in front of the API server, and infrastructure defined in code instead of a pile of shell scripts I'd forget how to run in a month. This post walks through the shape of the setup.

Why kubeadm instead of k3s or minikube

Managed dev clusters hide exactly the parts I wanted to practice: certificate rotation, etcd topology, and how the control plane behaves when a node disappears. kubeadm keeps all of that visible.

The Terraform layer

Multipass VMs are provisioned declaratively, with cloud-init handling the base packages so every node starts from the same known state. The pattern I settled on:

• A modules/vm module parameterized by role (control-plane / worker)
• Remote state kept local for now, with a documented path to a remote backend
• Outputs feeding directly into the Ansible-free kubeadm init / kubeadm join steps

What I'd change next

The biggest gap right now is automated etcd backups and a proper tflint / Terratest pass before I call this "production-grade" rather than "production-shaped." Both are next on the list.