Search code examples
aeron

Is it possible to reduce the timeout (10s?) during cluster update by actively shutting down the leader?


If AeronCluster wants to update the leader node when running, the cluster will face the heartbeat timeout (10s?) unavailability. Is it possible to add a signal to ClusterTools to make the leader node abdicate and let other followers actively start the election? (I tried remove-member, but the node cannot join the cluster later.


Solution

  • This is possible by running the media driver separate from the consensus module on each node. To reduce the 10s timeout on the leader the consensus module on the leader can be closed while leaving the media driver on the leader running. The followers will then detect the log channel has closed and immediately enter a new election.