Hello Airflow Dev community,

We saw that you recently (Airflow 2.1.3) added a feature for having Queued
state for DAGRuns

https://github.com/apache/airflow/pull/16401

However we have noticed the two following behaviors:
* start_date does not get populated by default for the DAG Runs triggered
manually (either UI or using local_client.py for instance). Reason is that
the field is not specified in the create_dagrun call
(api/common/experimental/trigger_dag.py) and that now() got removed as
default value in the ORM definition (models/dagrun.py, but this makes sense
since dags are queued by default for the scheduled case)
* In the UI if you hover on a DAG Run (which was triggered manually) it
will print (Dag has not started yet), even though the DAG has succeeded
since the start_date is NULL in the DB

Therefore we would like to know whether this behavior is intended or if it
is a bug ? If intended could you please let us know what was the reason
behind this? In my opinion, start date should be filled even for manually
triggered dag as it is the "physical" date of the execution as opposed to
the execution_date which is a more logical one, and it is a useful piece of
information

Thank you for your support!
Regards,
Philippe

Reply via email to