fshehadeh commented on issue #56446:
URL: https://github.com/apache/airflow/issues/56446#issuecomment-3408459563

   Thank @potiuk for sharing these additional details. In our setup we run 
Airflow from container images. In production, we lock down the DAG files 
between releases, and in lower envs we update the DAG frequently. The one 
minute reparsing was aggressive for us, so by adding the checksum to it, we are 
able to skip the unnecessary reparsing, while still being able to detect 
changes within a minute (because the checksum would have changed).
   
   We also ended up adding an upper limit for how long the DAG can go without 
be reparsed (I think like 10 minutes). We were worried about possible negative 
side-effects of skipping the reparsing for a long time. Your comment about the 
callbacks and their relationship with the parsing logic explained why this was 
necessary. I think ideally the process of parsing should be decoupled from that 
of the callbacks.
   
   I think the approach can help some users, but not necessarily all of them, 
depending on the number of the DAGs, their frequency, and complexity.


-- 
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.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to