etcd-io/etcd

etcd is a distributed key-value store, essentially a database that stores data as simple name-value pairs (for example, "mykey" = "some value"), but designed to remain consistent and reliable even when spread across multiple servers. It is meant for the most critical configuration and coordination data in distributed systems: the kind of data that, if corrupted or lost, would break everything depending on it. The key technical property that makes etcd special is its use of the Raft consensus algorithm. Raft is a protocol that ensures all copies of the data across multiple servers agree on every change, even if some servers go offline or messages are delayed. A cluster of three etcd servers, for example, can keep working even if one server fails, because the other two can still reach agreement. This makes etcd "highly available", unlikely to go down during normal failure scenarios. etcd exposes a straightforward API, either through a command-line tool called etcdctl or through a gRPC network protocol (a high-performance communication standard), where you read, write, watch, and delete keys. The watch feature is particularly powerful: applications can subscribe to a key and be instantly notified when its value changes, which makes etcd useful as a coordination layer between distributed services. The most widely known use of etcd is as the backbone of Kubernetes, the container orchestration platform. Kubernetes uses etcd to store all cluster state: which nodes exist, which containers are running, what configurations are active. Virtually every Kubernetes cluster in production depends on etcd. You would use etcd when you need different parts of a distributed system, multiple servers or services, to share configuration, leader election (deciding which server is in charge), or distributed locks. The tech stack is Go, communicating over gRPC, with TLS encryption for security.