zjncs opened a new pull request, #4592:
URL: https://github.com/apache/rocketmq-dashboard/pull/4592
### Which Issue(s) This PR Fixes
- Fixes #<issue-id>
### Brief Description
`scanDeadLetters` abandons a DLQ queue mid-scan in two places without
incrementing `failedQueueCount`:
1. a pull whose `nextBeginOffset` does not advance (a stall that would
otherwise loop forever), and
2. more than `MAX_CONSECUTIVE_OFFSET_ILLEGAL` consecutive `OFFSET_ILLEGAL`
responses (giving up on the corrected offsets).
Only the exception path and the null pull result counted. Because
`scanIncomplete()` is `failedQueueCount > 0 || truncated`, a scan where every
*other* queue succeeded stayed "complete": the resend was classified as
`SUCCESS` with `scanIncomplete=false, scanFailedQueues=0`, even though part of
the dead-letter queue was never read. The operator gets a "complete" resend
whose `matched` count silently excludes the unscanned messages — for dead
letters that is the worst kind of success, because the unresent messages look
handled.
This PR adds `failedQueueCount++` before both abandonments, so:
- mixed case (one queue abandoned, others scanned): outcome becomes
`PARTIAL` with `scanIncomplete=true` and `scanFailedQueues=1`, matching the
existing semantics of the exception path
(`resendMessagesMarksAResultPartialWhenOneDlqQueueCannotBeScanned`);
- all-queues-abandoned case: the scan now fails with the same 502 `Failed to
scan DLQ topic ...` as the all-queues-threw case, instead of returning an empty
success.
The DLQ **list** path (`listMessages`) shares `scanDeadLetters` and already
maps `failedQueueCount` into its VO, so it picks up the corrected
incompleteness signal for free.
### How Did You Test This Change?
Two new tests plus one updated test in `RocketMQDLQProviderTest`:
- `resendMessagesMarksAResultPartialWhenOneDlqQueueStalls` (new): two queues
— one stalls (`nextBeginOffset == requestedOffset`), one yields an in-window
dead letter that is resent. Expects `matched=1, resent=1, outcome=PARTIAL,
scanIncomplete=true, failedQueueCount=1` and the `scanFailedQueues=1`/`PARTIAL`
audit record.
- `resendMessagesCountsAQueueAbandonedAfterPersistentOffsetIllegal` (new):
every pull returns `OFFSET_ILLEGAL` with a corrected offset that is itself
illegal again until the cap is hit. Expects the 502 `Failed to scan DLQ topic`
(the queue is the only one, so the whole scan fails) and no producer invocation.
- `resendMessagesStopsWhenPullOffsetDoesNotAdvance` (updated): the single
stalled queue now fails the whole scan instead of returning an empty success;
the test's original guarantees (exactly one pull, no producer, bounded time)
are kept.
Fail-before (fix stashed, tests on master's code): the three tests above
fail — the two-queue case returns
`SUCCESS/scanIncomplete=false/failedQueueCount=0`, the give-up case returns
instead of throwing. Pass-after:
```
$ mvn -B -ntp test
-Dtest='RocketMQDLQProviderTest,DLQServiceTest,DLQControllerTest'
-Dspotbugs.skip=true
Tests run: 41 ... in RocketMQDLQProviderTest (0 failures)
Tests run: 24 ... in DLQControllerTest (0 failures)
Tests run: 17 ... in DLQServiceTest (0 failures)
BUILD SUCCESS
```
The existing `resendMessagesRetriesFromCorrectedOffsetAfterOffsetIllegal`
(single illegal, then FOUND → `SUCCESS`) is unchanged and green, confirming the
retry-from-corrected-offset behaviour is untouched.
### Checklist
- [x] One coherent change; unrelated modifications are not bundled in
- [x] Commit subject follows Conventional Commits (`feat:` / `fix:` /
`refactor:` / `chore:` / `docs:` / `perf:`)
- [x] Tests added or updated for non-trivial changes, test methods named
`...Test`
- [x] New UI text has both Chinese and English entries under `web/src/i18n/`
(no UI change)
- [x] Architecture constraints stay green (`mvn test` runs the ArchUnit
checks)
- [x] New source files carry the ASF license header (no new files)
- [x] Documentation touched where behaviour changed (error/signal semantics
only, no user-facing docs exist for scan internals)
--
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]
