I’ve written a short article on how Kubernetes uses etcd, but the TL;DR is that Kubernetes uses etcd like a normal application uses a database.

It stores configuration data, state, and metadata in etcd. But – because Kubernetes is a distributed system that runs on one master and several worker nodes – that database needs to be distributed. etcd is a distributed database!

etcd has another feature that’s really useful to Kubernetes. etcd makes it easy to watch for changes to data stored inside it. This is useful because parts of Kubernetes are implemented as a control loop.

Kubernetes stores the ideal state and the actual state. If the control loop sees that these two states have diverged, it’ll make changes to reconcile them. The watch feature makes it easier to find out when the states have changed.