SEZ9 commented on issue #13052:
URL: https://github.com/apache/seatunnel/issues/13052#issuecomment-6841672431

   @KaustAbhinand thanks for confirming — that settles it. No further action is 
needed from you here; this issue stays with @Vivek1106-04's fix, and please 
don't open a competing implementation.
   
   @Vivek1106-04 to recap where this stands and what I'm waiting on:
   
   - Your follow-up commit `f1c125f62` on `fix-coordinator-error-generation` is 
stacked on the earlier #12165 head `a932e27a7`, so it can't be reviewed or 
merged on its own yet. Once #12165 merges, please rebase onto the current `dev` 
and open it as a focused PR linked to #13052; let's keep the review on that PR 
rather than on the fork branch.
   - In that PR, please show that each deferred path (trigger, completion, 
barrier, checkpoint-id) carries the generation that started the work and cannot 
update `errorByPhysicalVertex` or fail a replacement generation after 
`restoreCoordinator()`.
   - Keep `notifyCompleted`, `reportCheckpointErrorFromTask` and the 
task-running-failed path unchanged unless the PR demonstrates an unambiguous 
stale-generation path for each; a short note on why they were left on the old 
path is enough otherwise.
   - For the 4 new `CheckpointCoordinatorTest` cases, I'd like to see one that 
deterministically releases an old async failure after restore (restored 
generation stays running and keeps its own error state) and one showing the 
same failure still fails its owner when no restore happens.
   - If the two small scheduler/test nits from the #12165 review are going into 
this follow-up, please list them in the PR description so the scope is explicit.
   
   If #12165 has already merged by the time you read this, go ahead and open 
the PR; otherwise just ping me when it's up and I'll review the rebased head 
and its CI then.
   
   <!-- streview-comment:2303 -->


-- 
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