SEZ9 commented on PR #12043:
URL: https://github.com/apache/seatunnel/pull/12043#issuecomment-6888932185

   Thanks for the follow-up on `05bd79095d63600de4260b36b5ff16b6dfc5c208`.
   
   The change to the worker-restore fixture from `c9cde9325ac` sounds right: 
pulling the existing per-job config out of the `Optional` returned by 
`ReflectionUtils` and failing explicitly when it is absent is cleaner than the 
direct cast, and it keeps the success and stall assertions intact without 
touching production timers or public API. Since no local compile, runtime or 
E2E validation was run, I'll wait for GitHub CI on this head before treating 
worker restore and configured fail-fast as verified—please ping once it has 
finished.
   
   That said, this update only covers the test fixture, and I can't see 
anything in the thread yet that addresses the previous review findings. Could 
you reply to each of them (fixed on this head, planned, or disagree with 
reasoning), specifically:
   
   - **PR12043-F1 (HIGH)** – `startRestoreProgressTracking` enabling tracking 
before cancelling the previously scheduled check, so a stale timeout task can 
fire against freshly reset state.
   - **PR12043-F2** – the single interval+timeout deadline starting at restore 
time even when tasks aren't ready, so slow-but-healthy restores can be flagged 
and fail-fast can loop restore→timeout→restore.
   - **PR12043-F3** – tracking being started from `restoreCoordinator`, which 
also runs on the first transition to RUNNING, so never-restored pipelines get 
"post-restore" stall warnings.
   - **PR12043-F4** – `cleanPendingCheckpoint` mutating restore-tracking state 
outside the coordinator lock and not stopping an already-running timeout task.
   - **PR12043-F5** – the 10 ms validation floor for `restore-progress-timeout` 
combined with `restore-progress-fail-fast: true`.
   - **PR12043-F6** – measuring the stall timeout with 
`Instant.now().toEpochMilli()` instead of a monotonic clock.
   - **PR12043-F7** – `markAllTasksReadyAfterRestore()` recording readiness 
before `notifyTaskStart()` is confirmed.
   - **PR12043-F8** – the new options being documented only in the 
separated-cluster guides, not the hybrid-cluster guide (en/zh).
   
   F1 is the one I'd want resolved before merge; for the others, a short note 
on how you intend to handle each is enough for now. If some of them are already 
fixed in this head, just point me at the relevant changes and I'll re-check.
   
   <!-- streview-comment:2363 -->


-- 
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]

Reply via email to