potiuk commented on pull request #14792:
URL: https://github.com/apache/airflow/pull/14792#issuecomment-799615845
> I think your first hypothesis is most likely -- the parsing is kicked off
by a background process, and with the introduction of parallel tests the load
average will be higher, so the parsing process will likely be slightly slower
to read the DAG off disk and parse it.
I looked at the code and I believe this also means that Hypothesis 2) might
come into place. I believe some SchedulerJob originated processes might still
be running when test finishes and they can make all the DB updates that can
either remove existing data there or update it.
I guess we can stabilize the tests by not only executing clean_db() in
tearDown but also we could register all SchedulerJobs and end them gracefully
in the tearDown. I believe for example this could be the reasons for
https://github.com/apache/airflow/issues/14773 . I will experiment a bit and
push a more complete fix.
----------------------------------------------------------------
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]
