etcd

etcd is the distributed key-value store that Kubernetes trusts with its cluster state. Raft consensus, strong consistency, and a watch API that notifies you when keys change. If Kubernetes depends on it for everything, it's probably reliable enough for you too.

If you need a strongly consistent config store or service discovery backend, etcd is proven at scale. Consul from HashiCorp bundles service discovery, health checking, and a KV store in one package — more features, more complexity. ZooKeeper is the legacy option that etcd replaced in many architectures. Commercially, AWS has Parameter Store and Secrets Manager for simpler config management.

The watch API is underrated. Subscribe to key changes and react in real-time — great for dynamic configuration, leader election, or distributed locks.

The catch: etcd wasn't designed for large data volumes. It's a config store, not a database. Keep values small (default limit is 1.5MB per request). Running etcd outside of Kubernetes means operating a distributed consensus cluster yourself, which requires understanding quorum, backup strategies, and cluster recovery. Most teams use it indirectly through Kubernetes without realizing it.