Ankurdeewan commented on issue #59779:
URL: https://github.com/apache/airflow/issues/59779#issuecomment-3689180235

   Looking at the current implementation, `AzureBatchOperator` submits the job 
and then spends the remaining of its execution synchronously polling Azure 
Batch. Both pool readiness and task completion are handled via repeated SDK 
GET/list calls wrapped in sleep loops (`wait_for_all_node_state()` and 
`wait_for_job_tasks_to_complete()`), with no callbacks, streams, or long-lived 
connections involved.
   
   That control flow is inherently deferrable and effectively matches the AWS 
Batch operator’s model prior to deferral. Refactoring this to submit once, 
defer to a trigger for async polling, and resume on a terminal state would 
remove unnecessary worker blocking while keeping behavior unchanged, with 
`deferrable=False` remaining the default for compatibility.
   
   I’m happy to take this on and implement it following the 
AWSBatchOperator/BatchJobTrigger pattern, unless there’s a different direction 
that would be preferred.


-- 
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