Actions
Subtask #18784
closedFeature #18765: rbd-mirror: add support for active/active daemon instances
rbd-mirror A/A: leader should track up/down rbd-mirror instances
% Done:
0%
Source:
Tags:
Backport:
Reviewed:
Affected Versions:
Pull request ID:
Description
After acquiring the lock, the leader should read the "rbd_mirror_instances" mapping into memory. When the leader sends out its periodic heartbeat, the clients ack the message. Use the client (global) id received in the response to track which peers are alive and which are unresponsive after X missed acks. If a peer is down (missed X heartbeats), its state should be updated to down in the mapping table, should be blacklisted, its "rbd_mirror_instance.<client instance id>" object should be deleted, and the mapping should be removed in the instances table.
Actions