danny0405 commented on issue #19301:
URL: https://github.com/apache/hudi/issues/19301#issuecomment-4988470491

   Thanks for confirming that #19237  is already applied. This looks different 
from the coordinator event-buffer loss fixed there.
   In Hudi 1.1.1, the default heartbeat expiry is approximately two minutes, 
while the first recommit shown here occurs more than five minutes after the 
original failure. Also, hoodie.clean.manual.batch.enable is not a recognized 
Apache Hudi configuration, so Flink async cleaning is likely still enabled.
   
   PR #19237 was merged on top of #18904 on master. Applying only #19237 to 
1.1.1 does not include #18904’s resilient heartbeat writer or its increase of 
hoodie.client.heartbeat.tolerable.misses from 2 to 10.
   Could you test with #18904 applied as well, or temporarily configure:
   
   ```xml
   'hoodie.client.heartbeat.tolerable.misses' = '10',
   'clean.async.enabled' = 'false'
   ```
   
   Please also provide logs matching Heartbeat expired, rollback, clean, and 
the affected instant, plus the corresponding rollback entries from 
.hoodie/timeline. That will confirm whether the cleaner removed the inflight 
instant before coordinator recovery.


-- 
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