We were able to fix this by adding -z option to rsync option on the mirror side as that's implicitly set on ours as well and seems to cause issues on CentOS6

Also, I'll note again that according to timestamps the mirror is almost a month old so it should be removed from the mirrorlist. I'll remove it from http://resources.ovirt.org/pub/yum-repo/mirrorlist by hand if you don't mind

Some troubleshooting info about the mirror: the logs on the pod doing checks show the following:

MainThread::ERROR::2018-03-06 17:25:35,286::544:_aggr_files) fatal exception fetching files from http://mirrors.ibiblio.org/ovirt/pub
Traceback (most recent call last):
File "/opt/rh/rh-python35/root/usr/bin/mirror_checker.py", line 515, in _aggr_files
self.status[url] = int(timestamp)
ValueError: invalid literal for int() with base 10: '--2018-02-27 13:39:00-- http://plain.resources.ovirt.org/pub/src/.timestamp\nResolving plain.resources.ovirt.org... 2610:28:3091:3001:ffff:0:42bb:e61c, 66.187.230.28\nConnecting to plain.resources.o

This happenes in this line of mirror_checker.py while looping over these directories.

Results:

So one of the timestamps is garbage on the mirror (possibly due to some bug in mirrorchecker) and due to rsync issues it stays like this. I will continue investigating with the mirror admin to get it back to sync

is it enough to manually remove the mirror from http://resources.ovirt.org/pub/yum-repo/mirrorlist to ensure yum clients don't use it? I don't want to disable monitoring as this is just a temporary action.