fm-orchestrator

Clone
Source Code
GIT

#584 MBS appears to be getting buildroot/srpm-buildroot profiles from dist-git rather than PDC
Closed: Invalid 6 years ago Opened 6 years ago by psabata.

Closed: Invalid
Reopen Issue

I cannot tell whether this is really the case but today we ran into an interesting issue.

  1. We updated buildroot and srpm-buildroot profiles in base-runtime:master, adding fedora-modular-repos; we didn't build anything at this point.
  2. Shortly after @frostyx began complaining his COPR builds using the Rawhide modular compose started failing because fedora-modular-release couldn't be found.
  3. He tried building a package locally with mbs-build local and it failed with the same issue -- no fedora-modular-repos.

Note there were no base-runtime builds with fedora-modular-repos anywhere in them -- be it components, API or profiles. It seems as if MBS was using whatever was in dist-git rather than whatever was associated with that particular built module by quering PDC for it.

I don't know whether this affects the koji backend but it most likely does. I can test that by adding a bogus package to one or both of the profiles, push but not build and then attempt to build testmodule against that stream of base-runtime, if necessary.

It takes the dependencies and buildroot packages list from the last active module build in PDC.

https://pagure.io/fm-orchestrator/blob/master/f/module_build_service/builder/base.py#_279

Do you have MBS logs from that situation? Can you reproduce it on prod MBS?

What is more likely to happen is that we have some place in code which queries the PDC without the active=True, so it took the building module as a dependency. But I don't see any place like that right now. Logs would tell us :).

I was going to say the module wasn't even building, that the change was only in dist-git...

But I see it was, actually. I had no idea the change author submitted that particular change for build. Well, that explains everything. I'm closing this report. Sorry for that.

Metadata Update from @psabata:
- Issue close_status updated to: Invalid
- Issue status updated to: Closed (was: Open)
6 years ago

I'm betting that was me. After making the bad changes I was trying to figure out the mbs-build format. I thought that build didn't work because it was trying to authenticate via a browser, and I was using a virtual machine that didn't have one, when I was first trying. Evidently I got it going enough to build a bad build.
I'm sorry about that, and sorry I didn't say anything earlier. It was only overnight that I realized that what I thought was a failed build, must have gone through enough to mess things up.

Login to comment on this ticket.

Metadata
None
None
None
None
Powered by Pagure 5.13.3
Documentation • File an Issue • About • SSH Hostkey/Fingerprint
© Red Hat, Inc. and others.