Description =========== First, this *only* happens to ppc64le. Not noticed on x86-64/aarch64 builders, even under shortage/queue conditions. * Many (if not all) ppc64le builds systematically fails. * See some examples here, (builds was canceled) for exposure: $ copr list-builds rezso/HDL | grep cancel | awk '{print $1}' 5119900, 5119899, 5119882, 5119868, 5119855 5116866, 5116863, 5116862, 5116861, 5116860 5116859, 5116857 * Failures *always* end up with some kind of timeout: "Connection timed out during banner exchange" "Connection to XX.XX.XX.XX port 22 timed out" E.g. https://download.copr.fedorainfracloud.org/results/rezso/HDL/fedora-rawhide-ppc64le/05119899-verible/builder-live.log.gz Ocurrence: ========== * Before latest scheduled upgrade there was only few (waivable) failures. * After latest scheduled upgrade these failures are frequently occurring. Sometimes might be a shortage/queue of ppc64le builders, but this don't explain such failures.
Update ======= * Another strange behaviour, related to ppc64le only, that might hint at firewall/badcfg issues, despite "internet access" is checked: <...> Cloning into '.'... fatal: unable to access 'https://github.com/chipsalliance/Surelog.git/': Failed to connect to github.com port 443 after 1039 ms: No route to host <...>
Hello Balint, thank you for the report. I think it will be related to this issue https://github.com/fedora-copr/copr/issues/2433 It is my priority for this week
Sorry, it took so long, I had a hard time debugging this and needed some help from @praiskup. It should be fixed now, more information in the GitHub issue above. Please let us know if you see this happening again.