123123213weqw opened a new pull request, #6389:
URL: https://github.com/apache/rocketmq-dashboard/pull/6389
@lizhimins Supersedes #5754 (GitHub refuses to reopen maintainer-closed PRs,
hence the fresh one) — same paging fix, with the three review asks addressed,
and the branch merged onto the current `rocketmq-studio`.
## What
Attaching to a live run replayed only the first timeline page; a backlog
beyond one `DEFAULT_TIMELINE_LIMIT` slice never reached the attaching client.
This pages the replay forward under a cursor until the backlog is drained,
bounded by `MAX_REPLAY_PAGES` (25).
## Review asks from #5754, addressed
1. **Ordering vs #5753:** understood — this must land after #5753 (still
open), since on its own the loop widens the attach gap that change closes.
whenever you take #5753, this one follows immediately after.
2. **Silent truncation:** the ceiling is now loud — when the loop runs out
of pages with every page full, it logs `attach replay of run {} hit the {} page
ceiling after {} rows; older events are not replayed` at WARN.
3. **Cursor monotonicity:** confirmed — `findByConversationIdAfterSeq` has
no SQL `ORDER BY` on purpose (MEDIUMTEXT sort cost), but it sorts the bounded
slice in memory by `seq` then `id` (`MybatisPlusAiEventRepository.java:64-71`),
so `cursor = row.getSeq()` strictly advances and no page can be re-fetched or
skipped.
The regression test still goes red without the loop. `AiRunServiceTest` is
green on this branch.
--
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]
