11:26 <blu> hey people, any idea why a build-kernel job might be failing with an container image pull error: 'ERROR: Job failed: failed to pull image "registry.freedesktop.org/blu/linux/debian/bookworm-slim:2023-06-10.4" with specified policies [if-not-present]: initializing source docker://registry.freedesktop.org/blu/linux/debian/bookworm-slim:2023-06-10.4: reading manifest 2023-06-10.4 in registry.freedesktop.org/blu/linux/debian/bookworm-slim: manifest unknown:

11:26 <blu> manifest unknown (manager.go:251:0s)'

11:33 <bentiss> blu: which job please?

11:34 <blu> https://gitlab.freedesktop.org/blu/linux/-/jobs/71462592

11:34 <bentiss> (FWIW, I'm trying to purge the registry with old images, so you probably have to restart the job that pulled that image in your container registry)

11:35 <bentiss> blu: you should restart https://gitlab.freedesktop.org/blu/linux/-/jobs/66974175

11:36 <bentiss> but I don't understand why this 3 months old pipeline is running a job only now

11:36 <bentiss> (and yep, that fixed it, now the image is in your repo)

11:39 <blu> bentiss, thanks, that fixed it indeed. re the old job -- I just manually re-ran it.

11:39 <bentiss> k

11:40 <bentiss> well, I'm going to enfore a policy where images that are just a copy of the upstream manifest will be deleted after 7 days. This will allow our gc to run more efficiently. Just so you know

11:40 <bentiss> so you might need to re-run those container build jobs

11:41 <blu> bentiss, roger that. thanks for the heads-up!

11:59 <bentiss> daniels: so it seems that using a blank fastlky configuration is not ideal: lots of pages are not refreshed properly (whot gave up today). However, gitlab has the ability to configure a CDN for static objects only: https://docs.gitlab.com/administration/static_objects_external_storage/#set-up-external-storage and https://gitlab.com/gitlab-org/charts/gitlab/-/merge_requests/2559.

11:59 <bentiss> However, that means we need to work on a "worker" in fastly to conver those requests and add the proper headers

12:01 <bentiss> IOW, not sure I'll have the time to work on that

12:01 <bentiss> (good thing that we got a couple of volunteers)

12:18 <bentiss> gitlab security upgrade in progress (minor version so should be just a restart of all pods)

12:25 <svuorela> hah. I just hit that...

12:25 <svuorela> (the restart)

19:29 <svuorela> Some of the poppler CI jobs does quite a bit of 'fetching random packages from the interweb and install them' - can that somehow be cached at the f.do level rather than needing to be re-done on each build ?

21:19 <__tim> svuorela, yes, by using fdo ci-templates to build custom docker images, and then re-using those for the jobs