Hi ,
We have faced the similar issue and debugged the scheduler code and found
that back fill don't have much options given( Back fill logic is different
in more scheduling point) .In this case once the DAG is stuck it is stuck
. Instead of using backfill you can use trigger_dag which has more l
I was doing backfill running a one task dag. There were over 1000 dag_runs
ready to run, I set concurrency=16 and max_active_run=16, I use Celery executor
and have 12 workers. Task instances were quickly scheduled, but they stuck
there and never get run. I did a search online and got this: [jira