JkSelf commented on PR #52039:
URL: https://github.com/apache/spark/pull/52039#issuecomment-3199024408

   > > This PR ensures that the task is properly canceled ...
   > 
   > How is this possible? Driver can only send an RPC to kill a task, which is 
not guaranteed to success. The only reliable way to is to shut down the 
corresponding executors forciably, which is the job of task reaper 
([SPARK-18761](https://issues.apache.org/jira/browse/SPARK-18761)).
   
   @cloud-fan Thanks for your review.
   
   I tried removing 
`sparkSession.sparkContext.cancelJobsWithTag(executionIdJobTag(sparkSession, 
executionId))` and directly setting `spark.task.reaper.enabled` to `true`, but 
it doesn't seem to take effect. My understanding is that we still need to call 
`cancelJobsWithTag `to mark the task as killed, and then enable 
`spark.task.reaper.enabled` to forcibly kill the corresponding executor to 
ensure the task is eventually terminated. Is that correct?"


-- 
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