Cluster
Committed coordination view: registered brokers, queue ownership with fencing epochs, and the embedded consensus group when one is active. Solid edges point from queue owners to followers.
Drain this broker
Announces a planned restart to connected clients and, in coordinated mode, hands partition ownership to caught-up followers before returning. Zero partitions remaining means stopping the process is gap-free.
Summary
loading…
Recovery checkpoints
Agreed state boundaries shorten recovery history. Local retained ranges may include live payloads; disabling new checkpoints keeps accepted retention pinned.
Recovery timeline
Recent recovery worker attempts on this node. Timing starts when the worker begins; failure detection and client reconnection are outside this window. Concurrent stages overlap. Completed means the attempt returned successfully; replica admission can continue on retry.
History is held in memory and resets on restart. Inspect other nodes for their attempts.
Brokers
Hover a broker to highlight its replication links, or click to inspect its assignments. The consensus leader is marked.
Click a broker to inspect its assignments.
Assignments
Queue partition ownership from the committed snapshot.
| Topic | Part | Group | Owner | Followers | Epoch |
|---|
Repartition queue
Set a partitioned queue's partition count. The controller assigns owners and followers for the new layout.
Coordination membership
Add or remove a voting member of the consensus group. Quorum changes affect availability; change one member at a time.