Adjusting back towards Kubernetes

No plan survives first contact with the enemy

Previously I wrote about how I planned to uplift my self hosting setup to a new standard, implementing new tooling and strategies to enable me to enjoy more of my hobby. That effort is still ongoing however, I’ve felt I needed to change strategies on the fly. Indeed, I wrote about how I wanted to uplift my infrastructure layer to use LXD as a hypervisor along with LXC to enable a transition away from Docker and Docker Compose. I am still working on that transition, but instead the final destination is to my first tech love, Kubernetes.

When it comes to self hosting, you tend to find a lot of trepidation, or disgust, at the thought of running Kubernetes in the home. Certainly, were there a need for a picture to describe overkill, running Kubernetes in your closet is probably rather apt. There’s certainly solutions out there for it, such as lightweight distributions that strip out “unnecessary” components. Or you can go the full 90, which is an exceptional learning experience I imagine. Choosing Kubernetes for my self hosting environment was informed by trying to learn more but to also maintain my skills. I have not had to maintain production Kubernetes clusters at scale since 2021 when I changed jobs. I went from an administrator to a consumer in terms of my role. So, I have a basic desire to keep the blade sharp in terms of skill so to speak.

However I don’t wish to take on all the burden. Managing a control plane, even if it’s some lightweight distro, is not really my appetite currently. A lot of my experience is using managed cloud providers such as Google Cloud and Azure flavours of Kubernetes. Which is great for just letting you click buttons to get things done and you rely on the engineering hours of great engineers to hopefully, not screw up. Couple this with wanting a reliable foundation for my self hosting services, I wanted to figure out a best of both worlds.

I came across Scaleway Kosmos in my research. Already in the enterprise there is products where the idea is you basically connect your on prem compute as capacity for a Kubernetes cluster(s). Gain the benefit of that foundation of engineering hours I described earlier while re-using your existing footprint. Unfortunately such sales teams probably don’t entertain discussions with some random guy from Ireland. Fortunately, Scaleway seem more amenable to someone spinning a cluster up. Of course, there is a cost, approximately 105 euro per month at the time of writing. For the sake of engineering hours and my time, it comes across as a worthy investment.

So, I signed up for a Scaleway account, created a Kosmos cluster and began to research connecting all my local compute to the cluster. My mind was racing with all the possibilities of having a managed control plane and my own compute to power all my endeavours. Stay tuned for the next blog post, where I go over my experience in getting this all set up.