potiuk commented on issue #24037:
URL: https://github.com/apache/airflow/issues/24037#issuecomment-1143980447

   Surely - we can ask, but they are part of vmware so I have no high 
expectations their free offering will be looked at carefully. But I definitely 
will.
   
   However I think we should seriously consider moving it to Github (or 
vendoring in) regardless of their answers. What do we do if they answer "we 
fixed it" ? It happens intermittently so we are not even able to check  if they 
did.
   
   I really do not like situations where WE have to carry the burden of CI 
errors (and make our contributors unhappy) when someone else screwed up. 
   
   I am not sure if this is a serious effort - this is one-time effort really 
which needs zero maintenance (maybe upgrade from time to time). Comparing with 
multiple unforeseen errors that are yet another reson for our users to learn 
that "red" is normal and reaching out for help.  I think the key to keep CI 
"nice" for the users is to eliminate relentlessly any reasons for potential 
problems - leaving only those that are "real" problems as false negatives 
undermine your trust in it and the more it happens the more you reach out to 
maintainers "the tests failed, please help" - we should keep the possibility 
down to minimum. It can never be eliminated entirely but if we do not have 
control over something we cannot improve it.
   
   I've heard very similar concerns when I moved the images used during our 
integration tests. 
   And this worked beautifully after I did this for all the images we use in 
our integration tests  in ghcr.io :
   
   Do you remember the last time we had a problem with this image: 
https://github.com/orgs/apache/packages?tab=packages&q=airflow-openldap ? 
   
   Me neither. 
   
   And we had plenty of similar problems with those images from time time to 
time when they were pulled from dockerhub. It's literaly 0 overhead - I pushed 
it once 11 months ago and did not touch it since. I think the one time effort 
of moving stuff to GitHub (either as separate repo or vendor-in) is certainly 
worth it over the long period of time.
   
   The nice thing of keeping everything in GitHub is that when it fails, it 
generally fails in the way that many things do not work (including actions or 
CI in general). So when something breaks in GitHub people generally are aware 
of it (and more often than not they simply cannot even submit or run their CI 
jobs). 
   
   And if we see intermittent errors like that we have enterprise level support 
with them via Apache agreement - they either commented with workarounds or 
fixed many isues I raised to them already.
   
   
   


-- 
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