We have problems with official Fedora repositories accessibility in COPR builders. During last week, several people complained about failing builds because updates (or even base fedora) repo was not synchronized successfully. (even the worst case is probably when a build succeeds but with some unavailable repos, we need to counter that on the level of mock configs). Is there anything that can be done about this?
Example build where this happened is here: https://copr-be.cloud.fedoraproject.org/results/rpmsoftwaremanagement/rpm-gitoverlay-1515489494.289521/fedora-26-x86_64/00698466-dnf-plugins-extras/builder-live.log
Per my information from dnf team, this happens very often during last several days.
Similar error here: https://copr-be.cloud.fedoraproject.org/results/mcepl/vim8/fedora-rawhide-ppc64le/00695671-vim/root.log.gz
Here fedora could not have been synchronized from fedorainfracloud.org.
Is there anything that can be done about it? Are there dedicated Fedora mirrors inside infrastructure that we should use instead of public ones? Should we do that?
Could you get check_metalink -vv output from a COPR box and paste that in here please?
Here it is: https://paste.fedoraproject.org/paste/sdxTP0XxtyrUS~SdDwKMsg
Is the communication from builders to external network going through fedcloud09? Can fedcloud09 be checked for load (before there were network problems related to it).
Anyway, I will try to increase level of debugging info that dnf prints. That might help quite a bit, I think.
We have employed ChrootScan mocks' plugin get to dnf.librepo.log from buildchroot. Hopefully that might help us to get more information if the problem occurs again.
Should be now fixed. The problem was in proxy06 that was serving stale data.
Metadata Update from @clime: - Issue close_status updated to: Fixed - Issue status updated to: Closed (was: Open)