Pretext

In 2021, I purchased three identical Precision 3420 workstation towers, each with 32GB of RAM and a Xeon E3-1245 v5 CPU. The intention was to build some kind of cluster. Originally this was OpenStack; then oVirt. Eventually I gave up on orchestration and simply layered GlusterFS across the three systems for live migration in plain libvirt. By that time, I had spent so much time trying to produce any cluster at all that I lost all sense of what I could deploy. That persisted into my first attempt at Kubernetes in November 2025, however with my homelab now hosting half a dozen services, infrastructure is no longer a solution in search of a need.

The Plan

Over the years, I harvested disks from my cluster nodes for various purposes, but upon review today, I found three of the four severely underutilized. This presents an opportunity: I can reclaim two disks for immediate use in the second cluster node. Given this new capacity, I intend to deploy Ubuntu along with Microk8s. This will serve as a proof of concept, allowing me to migrate my existing services to Kubernetes; and - only when I am satisfied and ready - to reprovision the other two nodes, allowing them to join the cluster and enable high availability.

Why Microk8s?

In my first foray into Kubernetes, I found Microk8s by far the easiest to deploy. In reaching that conclusion, I taught myself something valuable: the point isn’t to learn how to deploy a Kubernetes cluster from scratch; the point is to learn operations, maintenance, and backup for an existing cluster. But beyond simple cluster creation, I can install Microceph for just as easy hyperconverged, highly available storage when I expand to the other two nodes.

Further Study

Alongside the proof of concept cluster, I’d like to acquire a copy of The Book of Kubernetes to study this subject more formally. I’m not sure that I’d ever pursue a career in devops or anything; but I consider it generally edifying to try to learn as much as I can, beyond simply beating my head into a computer chassis until the manifest deploys.

All in all, I’d say I’ve got some work ahead of me.