Il 27/09/23 23:49, Neal Gompa ha scritto: > On Wed, Sep 27, 2023 at 5:41 PM Kevin Fenzi <kevin@xxxxxxxxx> wrote: >> On Wed, Sep 27, 2023 at 10:05:38AM -0400, Neal Gompa wrote: >>> On Wed, Sep 27, 2023 at 9:48 AM Richard Hughes <hughsient@xxxxxxxxx> wrote: >>>> On Wed, 27 Sept 2023 at 13:23, Mattia Verga via devel >>>> <devel@xxxxxxxxxxxxxxxxxxxxxxx> wrote: >>>>> Can't this script be moved to run in Openshift as cron-based? >>>> Yes! In fact, that's what I proposed about a decade ago when I wanted >>>> to include the data in the metadata like Debian does. I do think it >>>> should be managed by someone in the rel-eng/infra team rather than me. >>>> >>> openSUSE and Mageia both do it this way too. If we move to running the >>> script in infra, we should just make the switch to appending it to the >>> repodata. When I helped Mageia set it up[1], we elected to compose it weekly >>> and re-append the results for every repo push until the appstream >>> repodata was regenerated. >>> >>> Also, our compose process *has* gotten faster over the past decade, so >>> we may be able to do this at compose-time now. >> Here's the old releng ticket about this (I think): >> https://pagure.io/releng/issue/5721 >> >> In addition to appstream data, there's the screenshots. >> >> I agree much has changed and we should revisit how best to do this >> process. >> >> I'm not sure at all that it would be possible to do at compose-time... >> composes are taking around 3.5-4hours and thats after I have done >> a lot to speed them up, but might be worth some benchmarking >> to see how much slower it would be. If we are going to go that route, we >> should see if support could be added to pungi. >> >> Moving this from a package to in repodata would grow the repodata quite >> a lot right? >> > It's additional repodata files, files that are not even downloaded by > anything except PackageKit (currently). So the main repodata sizes > don't change. > Perhaps we can mimic the bodhi-critpathcron pod and run the script within the bodhi openshift role once a week? I haven't looked in depth into the script, but I think we'll need to mount the archive from somewhere in network to avoid cloning the main mirror, run the script and then synch the output somewhere? I'd suggest to avoid adding the step within pungi or bodhi-composer: Richard said it takes about an hour per release and I don't think we need to have it run at each compose (it would be nice, though, but we're now running it once per month and the world is still here). Also, bodhi-composer already does too much things and when a compose fails a lot of garbage is left behind (broken synch between update states in bodhi and koji builds tags), so better not add another possible failure point. Mattia _______________________________________________ devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@xxxxxxxxxxxxxxxxxxxxxxx Do not reply to spam, report it: https://pagure.io/fedora-infrastructure/new_issue
