SEZ9 commented on PR #12678:
URL: https://github.com/apache/seatunnel/pull/12678#issuecomment-6962265784

   @kalayciburak thanks for the clear walkthrough of the behaviour at 
`71ac656652f0`. Here is where the two points stand.
   
   **Restart bound**
   
   Keeping restore seeding in #12094 is still the plan for this PR. The 
remaining ask is the description, which, as you noted, does not yet spell out 
the full impact. Please state explicitly that:
   - a new coordinator starts with an empty deque;
   - it does not select the files written before the restart;
   - after one restart the total can reach roughly 2x the configured limit, and 
repeated restarts can take it higher.
   
   **Delete failure**
   
   The requeue-to-front and retry on the next eligible completion that you 
describe addresses the concern in principle. As you said, 
`testRetentionDeleteFailureRetriesOnNextCompletion` stubs `notifyCompleted` and 
mocks the failures, so it is not yet proven end to end. Still needed:
   - The storage-backed test, showing that checkpoint completion succeeds after 
a reported `deleteCheckpoint` failure and that a later completion actually 
removes the failed ID.
   - A link to a Build run for the revision that contains that test. The 
currently linked Build check (122810327224) has no matching run, and refreshing 
dev will not create the fork run by itself.
   
   Once the description is updated and the test and its Build link are posted, 
I'll take another look at both points.
   
   <!-- streview-comment:2411 -->


-- 
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