Yeah, not a fan of any of the above approaches. We should rather update the Fedora Rawhide image each day, which should solve the problem also. Tried to do that now. If that works out, I will work on automating the process.

I hope there are new Rawhide composes every day. I'd be afraid that Rawhide composes fail quite often due to various dependency problems. But hopefully not.

But what if the image update fails for some reason? Then you'll (temporarily) break many tests with no easy workaround.

Not to mention that there will be some window between repo update and image update where the issue will still happen.

Yep, it worked \o/

@omos we are updating the to the latest image for few months now, did not yet break. Anyway, it is expected of the Rawhide images to get gradually better over time.

Anyway, we can add the current kernel packages as a separate repository via ansible playbooks, if we need to do this thing, this seems to be the most reasonable approach from my perspective

It happened here again:
https://src.fedoraproject.org/tests/selinux/pull-request/126

@mvadkert would something like this work (?):

• take the latest compose
• dnf update-it
• restart
• run tests

@mvadkert would something like this work (?):

• take the latest compose
• dnf update-it
• restart
• run tests

Ideally I think this wouldn't be a problem, but at this point the update of some unrelated build could break the system.

BTW, the prior Fedora CI also had this issue initially and they resolved it somehow. Unfortunately I can't find the ticket (if there was one) anymore...

I think this works on old pipeline, because it enables koji latest repo for the release:

https://github.com/CentOS-PaaS-SIG/upstream-fedora-pipeline/blob/master/config/Dockerfiles/fedoraci-runner/virt-customize/virt-customize.sh#L97

I managed to work around it by adding this to the playbook:
https://src.fedoraproject.org/fork/omos/tests/selinux/c/8d07664e720dd2d6c20ad89c3c9ade522241de1f

Happened again https://osci-jenkins-1.ci.fedoraproject.org/job/fedora-ci/job/dist-git-pipeline/job/master/37773/testReport/

This should be now resolved for Rawhide, we update to latest packages and reboot.

@plautrba @omos do we need this also for older Fedoras?

@mvadkert Are you sure that older Fedora images will be fresh enough to contain the latest stable kernels?

If it's the official release images (i.e. those that contain the kernel package from the time of release), then there should always be matching kernel-* packages available in the "fedora" repository. So for stable Fedoras the update&reboot shouldn't be required (unless the image is customized somehow).

it is the official ones. Ok so seems this is resolved. Closing.

it is the official ones. Ok so seems this is resolved. Closing.

it is the official ones.

@mvadkert It seems that this is no longer true? The F38 image currently boots with kernel 6.8.3, for which there are no matching packages in the repos:

https://artifacts.dev.testing-farm.io/7ba2c715-4d34-45dd-8a24-3a588cc43c4e/work-tests.ymlyoit87hq/tests-p3nv3r0m/ERROR-selinux-selinux-policy-hostapd-and-similar.log