@cverna Can you take a look at this?

I think this is related to that https://pagure.io/fedora-infrastructure/issue/6741

Not sure we can do much and that does not seems to happen very often.

I wouldn't say it does not happen very often -- in the case above, it happened 6 times in 7 attempts. I'm trying again with a different container image and so far have seen 3 failures in a row:

https://koji.fedoraproject.org/koji/taskinfo?taskID=44221862
https://koji.fedoraproject.org/koji/taskinfo?taskID=44222461
https://koji.fedoraproject.org/koji/taskinfo?taskID=44350654

Trying to overcome this with a cycle now:
while true ; do fedpkg container-build && break ; sleep 10m ; done

Am I just unlucky because I run the container builds at wrong time time of the day? Any idea when the master mirrors are updated, so I can try different time of the day, whether it is better?

So openQA used to run into this problem a lot. My general approach to work around it is to tweak the repository configuration to use mirrorlist instead of metalink. That way the infra repo always comes back at the top of the list and we always use it - all the clever-clever metadata timestamping stuff is skipped.

Would it be plausible to do that in this case?

So openQA used to run into this problem a lot. My general approach to work around it is to tweak the repository configuration to use mirrorlist instead of metalink. That way the infra repo always comes back at the top of the list and we always use it - all the clever-clever metadata timestamping stuff is skipped.

Would it be plausible to do that in this case?

Is this still happening?

I'm going to close this now... if it's still happening, please re-open. Thanks!