(2.12) Stream leader should propose consumer remaps once assignment processed #7071
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
d13b298 to
15f4c05
Compare
5399212 to
ee5277d
Compare
735ec1d to
9034b73
Compare
derekcollison
approved these changes
Jul 21, 2025
| @@ -29,6 +29,8 @@ import ( | |||
| "testing" | |||
| "time" | |||
|
|
|||
| "slices" | |||
There was a problem hiding this comment.
PLace with all other single name imports
Signed-off-by: Neil Twigg <neil@nats.io>
Signed-off-by: Neil Twigg <neil@nats.io>
9034b73 to
7928ef6
Compare
neilalexander
added a commit
that referenced
this pull request
Jul 29, 2025
The way that consumer assignments were proposed was changed in #7071 so that the stream leader is responsible for those updates instead of the metaleader. This subtly changes the timing involved in stream moves and cancels, as it is now not a given that the consumer assignments are always serialised with stream assignments in the same order. If a consumer assignment fails due to a cancelled stream move for example, we should not propose deleting the consumer, as the new stream leader will shortly come along and update the assignment correctly for the new placement. Signed-off-by: Neil Twigg <neil@nats.io>
derekcollison
added a commit
that referenced
this pull request
Sep 3, 2025
Revert #7071, since it would result in losing consumer state when scaling down a replicated stream and the consumer state would need to be moved, but the current stream leader doesn't host that consumer. `TestJetStreamClusterConsumerReplicasAfterScale` started flaking due to this bug. `TestJetStreamClusterConsumerReplicasAfterScaleMoveConsumer` has been added in this PR which specifically tests for this situation (the former test would only hit that randomly, hence it flaking). We'll need to revisit this later when implementing the stream-managed consumers, which isn't going to be for 2.12. Signed-off-by: Maurice van Veen <github@mauricevanveen.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Right now, in a stream scale up/scale down scenario where there are consumers that also need to be scaled, the metaleader makes a proposal to update the stream and then immediately makes proposals to update the consumers. However, this does not account for the fact that the stream proposal could fail to be applied, which is where we have sometimes seen peer set drifts.
This moves this logic to take place on the stream leader once the stream update proposal is applied instead.
This will also be necessary in a world where the consumer assignments are managed by the stream leader, as the metaleader will not know anything about the stream NRG.
Signed-off-by: Neil Twigg neil@nats.io