I thought we were still rebuilding/updating Cloud images for stable releases regularly. If not that seems like an issue; we really shouldn’t be allowing/encouraging people to deploy the day-of-release state of F33 to a cloud instance any more! Is this something we can address?
We build them (mainly because the the processes are set up to run nightly in releng), but they don’t go through any testing and none of our references to them get updated. This was one thing we did in Atomic Host was to work on automating the release (promoting tested Atomic Host) and updating the references (website), but it never landed in cloud, mostly because I didn’t have much time to work on it.
As far as I’m concerned the release day image is the only thing that’s ever guaranteed to work and as such I believe that it’s accurate to say that we don’t update them regularly. The artifacts are created, but that’s separate from the other bits (testing and updating links), which also need to be updated/automated if you want anything sustainable.
There are renewed efforts to release monthly or bi-monthly and fix these outstanding issues, but so far they haven’t landed.
one other point:
While I agree it’s not best practice I don’t really consider this any worse than someone using an F33 installer ISO to install F33 server/workstation. They just add a
dnf update and move on.
I think it’s probably a little bit worse because I expect more image-based use to expect to never update. So I think this is something we could improve. But yeah.
Well, they do go through openQA testing:
That’s a great start. We (fedora cloud group) used to have some QA box that would boot the qcow and vagrant boxes and run some basic tests. It worked well enough for basic sanity but we really needed more than that (and more than what we have with OpenQA today) to make sure that the images boot on the various clouds, which is something we really only got around release time by having people manually test it.
The good news is that we have some tooling in CoreOS land that we might be able to copy to do automated testing for Fedora Cloud too. We can get there, it’s just not where we’ve been in the past.
Very late reply, but if by “We (fedora cloud group) used to have some QA box that would boot the qcow and vagrant boxes and run some basic tests” you meant autocloud, those tests were actually migrated to openQA. The
cloud_autocloud test runs that same set of tests that autocloud used to run. See Tree - fedora-qa/os-autoinst-distri-fedora - Pagure.io for exactly what it does.