A distributed coordination service used for leader election, config management, and locks, the workhorse behind Kafka, HBase, and many others.
Apache ZooKeeper is a distributed coordination service that provides primitives for leader election, distributed locks, configuration storage, group membership, and synchronization. It runs as an odd-numbered cluster (3, 5, 7 nodes) and uses Paxos-like consensus for strong consistency. ZooKeeper has historically been the dependency of choice for systems that need rock-solid coordination: Kafka (until KRaft), HBase, Hadoop, Solr, and many internal tools. Newer designs increasingly use embedded Raft (etcd, Kafka KRaft) to avoid the operational burden of a separate ZK cluster.
Coordination is the hardest problem in distributed systems and ZooKeeper is the well-trodden path through it. The cost is real: ZK clusters have their own operational story (sizing, snapshots, transaction-log management, JVM tuning) and a misconfigured ZK takes down everything that depends on it. For new designs, evaluate whether embedded consensus (etcd, Raft libraries) is enough before committing to operating ZooKeeper.
See the part of the platform that handles zookeeper in production.