[Zen2] Fix test failures in diff-based publishing #35684
Conversation
`testIncompatibleDiffResendsFullState` sometimes makes a 2-node cluster and then partitions one of the nodes from the leader, which makes the leader stand down. Then when the partition is removed the cluster re-forms but does so by sending full cluster states, not diffs, causing the test to fail. Additionally `testDiffBasedPublishing` sometimes fails if a publication is delivered out-of-order, wiping out a fresher last-received cluster state with a less-fresh one. This change adds a freshness check to avoid this.
|
Pinging @elastic/es-distributed |
There was a problem hiding this comment.
I would like to avoid throwing a CoordinationStateRejectedException in PublicationTransportHandler. The validation logic should be left to CoordinationState. We can and should improve the logic on when to cache the incoming cluster state though, to avoid old cluster states from poisoning the cache. I wonder if the simplest way to do this would be to only store the new state in lastSeenClusterState after it passes the call to handlePublishRequest.apply. WDYT?
D'oh of course that makes much more sense. Somehow I missed that we were passing it to the coordinator here. I've done that instead. |
testIncompatibleDiffResendsFullStatesometimes makes a 2-node cluster andthen partitions one of the nodes from the leader, which makes the leader stand
down. Then when the partition is removed the cluster re-forms but does so by
sending full cluster states, not diffs, causing the test to fail.
Additionally
testDiffBasedPublishingsometimes fails if a publication isdelivered out-of-order, wiping out a fresher last-received cluster state with a
less-fresh one. This change adds a freshness check to avoid this.