runroot changing during the course of a rawhide pungi run

[Dusty Mabe]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256


If I understand correctly there was a new lorax build [1] that completed
at 01:11 UTC (02/23) that then made it into the runroot of a task [2] that 
was part of a pungi compose [3] that started at 14:54 UTC (02/22).

I think this caused problems with the build, but that isn't really
important. The real question is: if our runroot changes during the run then
we could have some tasks (lorax imagebuild etc) that run with some versions
of software and other tasks that run with others. This is probably especially
maddening when trying to debug why one failed when others didn't.

Is my understanding correct?

Dusty

[1] https://koji.fedoraproject.org/koji/buildinfo?buildID=1048851
[2] https://koji.fedoraproject.org/koji/taskinfo?taskID=25243479
[3] https://kojipkgs.fedoraproject.org/compose/rawhide/Fedora-Rawhide-2018022...
-----BEGIN PGP SIGNATURE-----
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=/fUr
-----END PGP SIGNATURE-----

[Kevin Fenzi]

On 02/22/2018 07:34 PM, Dusty Mabe wrote:
>  
>  If I understand correctly there was a new lorax build [1] that completed
>  at 01:11 UTC (02/23) that then made it into the runroot of a task [2] that
>  was part of a pungi compose [3] that started at 14:54 UTC (02/22). 
runroot is confusing here as we have a koji plugin called runroot that
runs tasks in a chroot. I think you mean buildroot here?

>  I think this caused problems with the build, but that isn't
> really
>  important. The real question is: if our runroot changes during the run then
>  we could have some tasks (lorax imagebuild etc) that run with some versions
>  of software and other tasks that run with others. This is probably especially
>  maddening when trying to debug why one failed when others didn't.
>  
>  Is my understanding correct? 
Yes, the rawhide buildroot is updated as soon as builds complete and are
signed (plus a small time for the newrepo to run).

This is not the case with any of the bodhi enabled releases. They only
update the buildroot from base packages + stable updates + specific
packages that are added as buildroot overrides.

kevin

[Dusty Mabe]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256



On 02/26/2018 07:51 PM, Kevin Fenzi wrote:
>  On 02/22/2018 07:34 PM, Dusty Mabe wrote:
> >
> > If I understand correctly there was a new lorax build [1] that completed
> > at 01:11 UTC (02/23) that then made it into the runroot of a task [2] that
> > was part of a pungi compose [3] that started at 14:54 UTC (02/22).
>  
>  runroot is confusing here as we have a koji plugin called runroot that
>  runs tasks in a chroot. I think you mean buildroot here? 
Yeah the buildroot (since that is where the runroot installs packages from).
 
>  
> > I think this caused problems with the build, but that isn't really
> > important. The real question is: if our runroot changes during the run then
> > we could have some tasks (lorax imagebuild etc) that run with some versions
> > of software and other tasks that run with others. This is probably especially
> > maddening when trying to debug why one failed when others didn't.
> >
> > Is my understanding correct?
>  
>  Yes, the rawhide buildroot is updated as soon as builds complete and are
>  signed (plus a small time for the newrepo to run). 
this is kind of unfortunate because a developer could torpedo a pungi run 8
hours in.

>  
>  This is not the case with any of the bodhi enabled releases. They only
>  update the buildroot from base packages + stable updates + specific
>  packages that are added as buildroot overrides. 
Yeah there is more control over bodhi, which is good. What about branched composes?

Dusty 
-----BEGIN PGP SIGNATURE-----
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=5UWC
-----END PGP SIGNATURE-----

[Kevin Fenzi]

On 02/26/2018 04:59 PM, Dusty Mabe wrote:
>  
>  
>  On 02/26/2018 07:51 PM, Kevin Fenzi wrote
> > This is not the case with any of the bodhi enabled releases. They only
> > update the buildroot from base packages + stable updates + specific
> > packages that are added as buildroot overrides.
>  
>  Yeah there is more control over bodhi, which is good. What about branched composes?
It depends on if bodhi is enabled for it yet. Right now, branched f28 is
like rawhide, next week when we hit bodhi enablement it will be like
stable releases.

kevin