amoghrajesh commented on issue #24171:
URL: https://github.com/apache/airflow/issues/24171#issuecomment-4574438540

   Same problem exists for spark on k8s. The `spark-submit` blocks for the full 
job duration watching the driver pod via `LoggingPodStatusWatcherImpl`, holding 
similar heap size while doing no actual work. Also blocks making the operator 
deferrable when/if needed.
   
   Planning to fix it with the same pattern as #65991 (yarn_track_via_rm_api): 
a `track_driver_via_k8s_api: bool = False` flag on 
`SparkSubmitHook/SparkSubmitOperator` that terminates `spark-submit` early 
after pod creation and polls via the K8s client instead. 


-- 
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