Hanging of GitLab CI jobs on a subset of job runners

Hello all,

Since 13th March one of GParted projects CI jobs has been failing if and only if it runs on one of the gcc runners. Before that date the CI job succeeded on all runners and it still does when it runs on a gnome (dot) org runner. The job is using a CentOS 7 docker image and trying to upgrade and install the needed packages when it seems to hang on a gcc runner, taking longer than the 1 hour timeout.

Example working job from 10th March

Example failed jobs from 13th March. Retried multiple times until it happens to run on a gnome (dot) org runner and succeed

I also noticed that when the gcc runners started failing is when the kernel reported newer version 5.16.18 (fedora) from the previous 5.13.0 (Ubuntu).

I’m a bit stuck and don’t really know what to do. Can anybody help and where should I report this?
Is there any chance of a solution, or should I just give up on a CentOS 7 image and use something newer?

(We like to be able have GParted compile and run on all supported distros, even the old ones, hence the use of this old image).


It looks like it has been caused by the mismatch between the number of file descriptors available on the host and inside the container. I’ve pushed a change to enforce the host one; https://gitlab.gnome.org/GNOME/gparted/-/jobs/2033337 passes yum update -y stage now.

Great. That fixed it.