RockteMQ-AI commented on issue #10658:
URL: https://github.com/apache/rocketmq/issues/10658#issuecomment-5067593897
**Issue Evaluation**
Category: `type/bug` | Status: **Confirmed** | Severity: **Medium**
### Verification Summary
Verified against `PopConsumerService.java` on `develop` branch. The
described head-of-line blocking behavior is confirmed in the code.
### Root Cause Analysis
Two failure paths in `revive(AtomicLong, int)` cause the entire batch to
abort:
1. **Synchronous throw (line 593-597):** When `this.revive(record)` throws
synchronously (e.g., `getMessageAsync` completes synchronously with a throw),
the `catch` block re-throws as `RuntimeException`, aborting the for-loop. All
subsequent records are not processed.
2. **Async exceptional completion (line 617):** `revive(PopConsumerRecord)`
at line 552 chains `getMessageAsync(record).thenCompose(...)` with **no
`.exceptionally()` handler**. If any future completes exceptionally,
`CompletableFuture.allOf(...).join()` throws `CompletionException`, skipping:
- `writeRecords(failureList)` — line 618
- `deleteRecords(consumerRecords)` — line 619
- `currentTime.set(...)` — line 620
This means healthy records in the same batch are never persisted as
processed, and the batch gets stuck retrying the same failing record
indefinitely.
### Scope
Confirmed: popkv path only (`popConsumerKVServiceEnable`), disabled by
default. Does **not** affect legacy `PopBufferMergeService`.
### Suggested Fix Direction
- Add `.exceptionally()` to the per-record future in
`revive(PopConsumerRecord)` to isolate failures
- Wrap the per-record call in the batch loop to catch both sync and async
failures
- Ensure `writeRecords`, `deleteRecords`, and `currentTime` advance always
execute even when individual records fail
---
*Automated evaluation by github-manager-bot*
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
