SEPURI-SAI-KRISHNA commented on PR #13261:
URL: https://github.com/apache/seatunnel/pull/13261#issuecomment-6985116629

   @DanielLeens Thanks, fixed in `6dfb56cf65`. The failure is now armed from a 
completed checkpoint:
   
   - The server answers every request at first. The test runs the job in the 
background and finds it by name in `/jobs/overview` of the Flink REST API (the 
test containers name a job after its config file).
   - It waits until the reader has read a row, reads the completed count from 
`/jobs/<id>/checkpoints`, and waits until two more checkpoints have completed. 
The job config sets `execution.max-concurrent-checkpoints = 1`, so the second 
of them was triggered after the reader read a row with its split.
   - Only then does it add a 500 expectation with a higher priority, and it 
asserts that the server sent exactly 3 rejections, one per attempt.
   
   Each wait is bounded (2 minutes, and 5 for the job). The search for 
"Completed checkpoint" in the logs and the fixed count of 8 answers are gone.
   
   On Flink 1.13, 1.15, 1.18 and 1.20 the test passes (4 of 4). Against the dev 
enumerator it fails on all four with `the enumerator failed when a split was 
handed back`: `SingleSplitEnumerator.addSplitsBack` throws when Flink hands the 
split back after the restart, so the test reproduces #13253 after a completed 
checkpoint.
   


-- 
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