Description of problem: Today, some 90% of my builds in mkyral/plasma-unstable fail shamefully due to problems when downloading build dependencies (sometimes even in syncing repos) [1]. That is very frustrating and it makes copr completely unusable. See for yourselves [2]: all of the failed builds of plasma-5.14 failed due to this issue. [1] https://copr-be.cloud.fedoraproject.org/results/mkyral/plasma-unstable/fedora-28-x86_64/00807908-kgamma/builder-live.log [2] https://copr.fedorainfracloud.org/coprs/mkyral/plasma-unstable/builds/ FYI: this kind of problems started to appear few months ago. It was hardly bearable when it screwed 10-20% of the builds. Today, it breaks almost all the builds. Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: 1. 2. 3. Actual results: Expected results: Additional info:
Sorry for the problems. We are looking at the problem with Fedora infrastructure. It was an error when downloading packages from public mirrors.
Hello, I filed an issue about this here where you can follow details: https://pagure.io/fedora-infrastructure/issue/7301 Could you, please, retry the builds if the problem persists?
New release of plasma shows the problem is still there. It behaves better, when I spawn the builds sequentially rather in parallel, which makes the build painfully slow - but even so around 30% of the builds fail the to the problem. See how serious it is: https://copr.fedorainfracloud.org/coprs/mkyral/plasma-unstable/builds/ All the fails on 5.14.1 and 5.14.0 (12 pages of them now) are caused by insfrastructure. Not one of them is actual build problem.
That's pretty bad. We will try to come up with some solution to this and keep you informed.
Hello Martin, I've opened Bug 1643281 - dnf does not try other mirrors if the current mirror fails when downloading a package and I've also pushed https://infrastructure.fedoraproject.org/cgit/ansible.git/commit/?id=dbb822036b9e4da778972b4a40fec2df0b8ca829 to help debug the issue. Could you do another try of the rebuild so that we can potentially collect more data? Thank you!
(In reply to clime from comment #5) > I've opened Bug 1643281 - dnf does not try other mirrors if the current > mirror fails when downloading a package These kind of errors are probably fixed already in the librepo package, dnf team confirmed that the related code got rewritten. Also I haven't seen those problems for a long time. I'm closing this bug. > and I've also pushed > https://infrastructure.fedoraproject.org/cgit/ansible.git/commit/ > ?id=dbb822036b9e4da778972b4a40fec2df0b8ca829 to help debug the issue. I reverted this, it unnecessarily overloaded the dl05 server (which isn't otherwise used in copr), and I don't really got the point of that change (except that it was deterministically broken either on all builders, or nowhere). There remains one problem with getting mirrorlist/metalink files, as discussed in: https://github.com/rpm-software-management/librepo/pull/158 https://github.com/rpm-software-management/librepo/pull/159 https://pagure.io/fedora-infrastructure/issue/7987 Which should be in copr production soon (TM). But this is orthogonal thing.