rm_stm/idempotency: fix the producer lock scope #16706
Merged
Conversation
This file contains 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
In case of a replication error of current sequence, the code issues a manual leader step down to prevent the subsequent requests making progress as that violates idempotency guarantees. This is done by holding a mutex while the request is in progress. The mutex is incorrectly released before issuing a step down in such cases which may theoretically let other requests make progress before step down is actually issued, the race sequence looks like this seq=5 replication_error seq=6 makes progress seq=5 issues a stepdown This bug was identified by just eyeballing the code but couldn't be verified due to lack of trace logs in many partitions test. Seems like something that should be tightened regardless. Deployed the patch on a 3 node cluster with 500MB/s OMB run, no noticeable perf changes.
|
Fixes #16657? |
| @@ -1182,6 +1181,7 @@ ss::future<result<kafka_result>> rm_stm::do_idempotent_replicate( | |||
| req_ptr->set_value<ret_t>(result.error()); | |||
| co_return result.error(); | |||
| } | |||
| units.return_all(); | |||
There was a problem hiding this comment.
it looks like this release is still to early as the step down happens on line 1129 outside of do_idempotent_replicate should we keep the units alive there ?
There was a problem hiding this comment.
I think it does? do_idempotent_replicate() takes a reference to the units as the input parameter.
There was a problem hiding this comment.
now i see it, it is a bit confusing
Thats the hope but I"m not 100% convinced (due to lack of diagnostics).. I can loop your test a few more times on this and see what happens. |
mmaslankaprv
approved these changes
Feb 27, 2024
|
/backport v23.3.x |
1 task
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.
In case of a replication error of current sequence, the code issues a manual leader step down to prevent the subsequent requests making progress as that violates idempotency guarantees.
This is done by holding a mutex while the request is in progress. The mutex is incorrectly released before issuing a step down in such cases which may theoretically let other requests make progress before step down is actually issued, the race sequence looks like this
seq=5 replication_error
seq=6 makes progress
seq=5 issues a stepdown
This bug was identified by just eyeballing the code but couldn't be verified due to lack of trace logs in many partitions test. Seems like something that should be tightened regardless.
Deployed the patch on a 3 node cluster with 500MB/s OMB run, no noticeable perf changes.
Backports Required
Release Notes
Bug Fixes