GH-4198: Fix async retry bookkeeping for Mono listener failures#4320
Merged
Conversation
9bd93da to
3e65a8e
Compare
artembilan
requested changes
Mar 5, 2026
artembilan
left a comment
Member
There was a problem hiding this comment.
Very good!
Please, add your name to the @author list of the affected classes.
Thank you!
Contributor
Author
|
@artembilan Thanks for the review 😊 |
artembilan
requested changes
Mar 6, 2026
… failures - Keep async failure records queued when retry is in progress - Remove only retry-target offsets from async-ack bookkeeping - Add regression test for Mono listener direct-throw retry behavior Signed-off-by: MinChul-Son <smc5236@naver.com>
…e processing Signed-off-by: MinChul-Son <smc5236@naver.com>
Signed-off-by: MinChul-Son <smc5236@naver.com>
* Drop unnecessary synchronized from async offset cleanup helper * Reuse retry-record lists and remove redundant list copies * Align removeOffsetsInBatch signature with handleRemaining wildcard record type * Simplify retry listener assertion in AsyncMonoDefaultErrorHandlerRetryTests Signed-off-by: MinChul-Son <smc5236@naver.com>
* Use common RecordInRetryException try/catch in invokeBatchErrorHandler and invokeErrorHandler * Keep single-record retry list creation with List.of(cRecord) * Add removeOffsetsInBatchForRetryRecords to verify offsetsInThisBatch/deferredOffsets cleanup * Remove IDE-specific SpringJavaInjectionPointsAutowiringInspection suppressions from async mono retry test Signed-off-by: MinChul-Son <smc5236@naver.com>
Contributor
Author
|
@artembilan Please take another look when you have a moment. |
artembilan
requested changes
Mar 6, 2026
artembilan
left a comment
Member
There was a problem hiding this comment.
Just couple questions in the test.
Thanks
Rename placeholder values, switch to typed mock() usage, and remove unnecessary poll sleep in removeOffsetsInBatchForRetryRecords(). Signed-off-by: MinChul-Son <smc5236@naver.com>
Contributor
Author
|
@artembilan I updated and applied request changes, Please re-review when you have time! |
Contributor
Author
|
Sry, I miss clicked |
artembilan
requested changes
Mar 9, 2026
- Use a snapshot-based pass for failed async records - simplify retry record initialization in invokeErrorHandler to avoid duplicate setup. - restore blank lines for readability Signed-off-by: MinChul-Son <smc5236@naver.com>
Signed-off-by: MinChul-Son <smc5236@naver.com>
artembilan
approved these changes
Mar 9, 2026
artembilan
left a comment
Member
There was a problem hiding this comment.
This is good now.
Will merge when build is green.
Thanks
Member
|
Merged and back-ported. Thank you for contribution; looking forward for more! |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
Fixes #4198
Summary
This PR fixes retry behavior when a @KafkaListener method has a Mono return type and throws an exception directly (before creating the async result).
Previously, with async acks enabled (default behavior for async return types), retry could stop after the first attempt because async-ack bookkeeping state could become inconsistent during RecordInRetryException handling.
This change keeps retry progressing while preserving unrelated async-ack state.
What was happening
In async failure paths, RecordInRetryException indicates the record should remain in retry flow.
However, container state handling around that path could lead to retry/pause issues.
Changes
Why this approach
A full async-ack state reset can affect unrelated in-flight records.
This PR removes only retry-target offsets to avoid collateral side effects while fixing the retry stall.
Tests
Validated with:
Scope
This PR addresses the issue path where the listener directly throws in a Mono return method.
It does not change semantics for Mono.error(...) completion-path handling.
Thanks for reviewing. I’m happy to adjust anything based on feedback.