potiuk commented on pull request #16653: URL: https://github.com/apache/airflow/pull/16653#issuecomment-868692975
Not sure about the origin of this, but I think it was a deliberate decision to run utcnow with every loop, because you get then slighly more correct when detecting an SLA - remember that time flows while the loop is running - so by the end it finished it could be couple of seconds but possibly even minutes) later - depending on how many tasks you have run in the last hour this loop might take quite some time. I think the way it was done was actually better. Now - even if this was a deliberate decision, I am not quite sure why ts (time at the beginning of the loop) is read and used in SLAMiss instead of the utcnow() (actual time now). Maybe there was a good reason why the "start" of SLA is used when we create SLAMiss .. Anyone can shed some light there? If anything, I'd rather use utcnow() also there. But maybe there are some good reasons for both choices? -- 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]
