SEZ9 commented on PR #12678:
URL: https://github.com/apache/seatunnel/pull/12678#issuecomment-6905495595

   Thanks for separating the two tests — that clears up what each one covers.
   
   On the retention delete failure point (PR12678-F1): 
`CheckpointCoordinatorTest#testRetentionDeleteFailureRetriesOnNextCompletion` 
covering both checked and unchecked delete failures with a requeue on the next 
completion is the right shape for the orphan concern. Since `notifyCompleted` 
is stubbed there and `CheckpointStorageRetentionPruneTest` only covers the 
successful on-disk prune, the one gap left is an end-to-end case: a real 
storage-backed `deleteCheckpoint` failure where the completion itself still 
succeeds and the ids are retried (and actually removed) on the following 
completion. Either extend the storage test with a failing-then-recovering 
storage, or point me at where that is already covered if I've missed it. Please 
also confirm in the PR description that a delete failure no longer fails the 
checkpoint.
   
   On the restore path (PR12678-F2): your latest comments don't touch this one 
yet. Is `completedCheckpointIds` now seeded from storage when the coordinator 
restores, so pre-restart files get pruned and storage stays bounded at 
`max-retained` after a pipeline restart? If so, a short test asserting the 
post-restart bound would close it; if not, that is still open.
   
   On CI: the local runs of `CheckpointCoordinatorTest` (18) and 
`CheckpointStorageRetentionPruneTest` (1) are noted and appreciated, but for 
head `71ac656652f0fa3423067dc6ee1052b3b3be844a` the 
[Build](https://github.com/apache/seatunnel/runs/122810327224) is still 
`ACTION_REQUIRED` with no fork run found. Please enable Actions on the fork / 
dispatch the existing Build workflow for this branch and link the run here — no 
extra commits or rebases needed for that.
   
   Summary of remaining asks:
   1. Storage-backed (non-stubbed) test for the delete-failure retry path, plus 
confirmation that completion isn't failed by a delete error.
   2. Answer/test for seeding `completedCheckpointIds` on restore.
   3. A fork Build run for the current head.
   
   <!-- streview-comment:2378 -->


-- 
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