znodes

In Apache ZooKeeper, znodes are the nodes that form the hierarchical namespace used to store small amounts of data and metadata. Each znode has a path, holds a data payload (as a byte array), and a stat with versioning information. The data payload is limited by a default size of 1 MB.

Znodes come in several flavors: persistent znodes remain until explicitly deleted; ephemeral znodes are tied to the client's session and are removed when the session ends; there are also sequential and ephemeral-sequential variants that append a monotonically increasing counter to the znode's name on creation.

Clients can read and write znode data, check existence, and enumerate children of a znode. The data and children can be protected by ACLs. A znode can have watches registered so clients are notified of changes to data or child lists; watches are one-time triggers.

ZooKeeper provides a linearizable consistency model and uses a protocol among a quorum of servers to ensure that changes are committed in order. Znondes are used for configuration management, naming, synchronization primitives such as leader election, distributed locking, and service discovery.

When a client session ends, all ephemeral znodes created during the session are deleted automatically. Changes to data or children increment version counters in the znode's stat, enabling clients to detect concurrent modification.