Goodbye my Kubernetes, Goodbye my friend

Helmsman Changing Course

I’ve ran some form of personal Kubernetes cluster for almost three years now. I got my first exposure to the technology in work and fell in love. So much so, that it became my focus in college via my final year project. Going into my graduate role, I would build and administer clusters for my day job. Throughout this time, my cluster moved from various providers, but it was always running a handful of workloads quite perfectly. Well, around two weeks ago now, I shut down my Kubernetes cluster, ending a three year run of cluster operations.

Why exactly? Well there’s a handful of reasons, for one, clusters are expensive. Even when you use a provider that does not charge you for the control plane, to run a cluster with redundancy you’re talking two, ideally three nodes. Depending on the spec, that can be 60-80 EUR in just compute alone. While I can certainly afford it, I never felt it was an entirely efficient use of my money because of my relatively low load factor. Not to mention a cost of time, once a cluster is running it’s generally low maintenance until you need to upgrade or something goes wrong. Your mileage may vary based on your provider, personal experience I think DigitalOcean’s offering was the most straight forward, even when compared to GKE for upgrades, but I didn’t want to think of it any more.

Another reason for me and the other big one, Kubernetes was very over kill for what I was doing. Ultimately there was around six to eight Pods present at any one time (excluding system level) and a few of those are just the replicas for the actual containers I’m trying to run. With all the other bits and pieces that go into a Kubernetes cluster, it had me wondering did I really need all that? And was it worthwhile to maybe go a different route if all I needed to do was just run a container or three?

Evidenced by the fact that I’m writing this post, I did find something that I felt was worthwhile to warrant a change. That will be discussed in my next post or two, the solutions aim was not to be cheaper (though it might be), but to just be overall more efficient and still give me a general ease around orchestrating the containers I needed to. I’ve been quite busy so I’ve not yet fully invested the time I need to in order to really get this new system going, but I must say it’s very promising early on, for a very little amount of time investment.