[GitHub] [airflow] ashb commented on issue #13542: Task stuck in "scheduled" or "queued" state, pool has all slots queued, nothing is executing

[GitBox <gi...@apache.org>]

ashb commented on issue #13542:
URL: https://github.com/apache/airflow/issues/13542#issuecomment-824886213


   @lukas-at-harren  Ahhhh, right, yes if you have dynamic dags and those dags disappear this is currently and edge case.  I've split that off as a separate issue, as I _think_ it's separate to the OP's issue.
   
   @Overbryd Do you have static dags (where _every_ time you parse a dag file it produces the same DAGs) or dynamic (where there is some loop/external file read where the result of parsing the dag file could produce different results)?
   
   What method do you use to get the DAGs available to the pods created by KubeExecutor? Is it possible that some pods are being created with "stale" dag definitions (if for example if you have DAG files baked in to the image and your image pull policy means that and old image is used on some nodes)?
   


-- 
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org