parthchandra commented on PR #57285:
URL: https://github.com/apache/spark/pull/57285#issuecomment-5123516698

   > Although it seems that @peter-toth 's comment 
(https://github.com/apache/spark/pull/57285/changes#r3639196967) is addressed. 
The `failureCount` + retry-wait handling resolves the no-retry problem, but on 
total failure the empty `Credentials` is still serialized and sent to executors 
before the retry fires (the Kerberos path skips the send entirely because the 
exception propagates). Worth either skipping the `schedulerRef.send` when all 
providers failed, or throwing so `updateTokensTask()`'s existing catch handles 
both the skip and the retry uniformly.
   
   Fixed. On total failure (all providers failed, nothing obtained) 
obtainTokensAndScheduleRenewal now throws instead of returning empty 
credentials, so updateTokensTask's existing catch uniformly skips the 
schedulerRef.send and schedules the retry — same as the Kerberos path. Partial 
success still sends the tokens it got and schedules on the successful 
providers' renewal time.
   Also added a new unit test for this.


-- 
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]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to