Around April of 2017 I wrote this article about moving my side project to a single-node Kubernetes cluster: https://hackernoon.com/lessons-learned-from-moving-my-side-project-to-kubernetes-c28161a16c69

As of today, the infrastructure is still running strong, although I’ve run into a few issues I will talk about later in this article. I initially set up my node as a $10/month node but it was barely not powerful enough. Since then, Digital Ocean seems to have roughly doubled the CPU/memory for each of its instances, so $10 might work out. The Kubernetes docs mention that your nodes should have at least 2 GB of RAM, which means a decent 3-node cluster would only run you about $30 a month. I am still on the $20/month node, but I’ve redeployed it a couple of times and resized it since the initial setup, for various reasons that I’ll go into here.

This title may already be ringing some faint alarm bells; why move a small side project to use Kubernetes? And was it worth it?

I run a small study site for tournament Scrabble players, available at https://www.aerolith.org. (Please excuse the lack of a nice landing page). It consists of a Django app, with PostgreSQL, and a small Go-based microservice that generates some complex word challenges on demand. The reason that service uses Go instead of Python had mainly to do with performance, but partially I was also experimenting with Go at the time. The site also uses SQLite to store the word databases, Gunicorn, Nginx, i.e. a standard Django setup. The site gets a couple hundred players a day, so it’s not that tiny. I have also used this platform to learn; the very first version of Aerolith for the web was completed in 2011 and written in ghastly Javascript. It is now a fairly modern app with React.js and ES6. So it’s a nice project because I get to tinker with technologies that I sometimes end up using at my day job (for example, I did some Redis/Sock.js stuff briefly in the past, as well as Django upgrades, etc).