ashb opened a new issue #16366:
URL: https://github.com/apache/airflow/issues/16366


   There have been various problems with the behaviour of max_active_runs not 
being respected (https://github.com/apache/airflow/issues/9975, 
https://github.com/apache/airflow/issues/14744, and we should try and fix this 
once and for all by changing how dag runs are created and run.
   
   Currently creating a DagRun in the webserver puts it immediately in to 
"running" state, which can end up with too many DagRuns being executed, causing 
problems for the scheduler.
   
   Clearing a DagRin can have similar problems.
   
   We should change the behaviour as follows:
   
   - If the webserver creates/clears and DagRun and would previously put it in 
to running state, it should put it in to "queued" instead
   - Make SchedulerJob._create_dagruns_for_dags _also_ create DRs in queued 
state.
   - Add a new stage to SchedulerJob._do_scheduling that checks For any queued 
DagRun if there is "space" to promote it to running
   - We can remove any checks about currently_active_runs inside 
_schedule_dag_run function
   


-- 
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:
[email protected]


Reply via email to