For the past few years I've spent most of my time building database operators on top of Kubernetes. Along the way I kept coming back to the same idea: under all the YAML and the moving parts, Kubernetes is really a framework for feedback controllers. It's control theory turned into something ordinary people can use. The same closed loop that keeps a thermostat at the right temperature is also the one that keeps your database alive.

So I finally sat down and wrote about it on the PlanetScale blog.

The post starts from running Postgres by hand, picking a node, attaching a disk, setting up replication, and writing a small watchdog script to fix things when they break. Then it walks through how each of those manual steps maps to a control loop inside Kubernetes, and why operators are built the way they are. It dips into a bit of control theory along the way. I think even if you don't use Kubernetes, you'll enjoy reading it.

You can read it here: The Feedback Loops Behind Kubernetes.