On Wed, Dec 21, 2016 at 10:55 AM, Sandro Bonazzola <sbonazzo@redhat.com> wrote:


On Wed, Dec 21, 2016 at 9:51 AM, Dan Kenigsberg <danken@redhat.com> wrote:
On Wed, Dec 21, 2016 at 10:17 AM, Sandro Bonazzola <sbonazzo@redhat.com> wrote:
> 00:00:31.874 Num Packages in Repos: 22534
> 00:00:31.875 package:
> ovirt-provider-ovn-1.0-1.20161219125609.git.el7.centos.noarch from
> check-custom-el7
> 00:00:31.876   unresolved deps:
> 00:00:31.876      python-openvswitch >= 0:2.6
> 00:00:31.876      openvswitch-ovn-central >= 0:2.6
> 00:00:31.876 package:
> ovirt-provider-ovn-driver-1.0-1.20161219125609.git.el7.centos.noarch from

It's good we have repoclosure, as it reminded us we cannot ship
ovirt-provider-ovn unless we build and ship a version of openvswitch
from from their master branch, at least until they ship ovs-2.7.

Sandro, Marcin: can we do it? Can we supply our own build of
openvswitch, like we did for Marcin's blog?

> check-custom-el7
> 00:00:31.876   unresolved deps:
> 00:00:31.876      python-openvswitch >= 0:2.6
> 00:00:31.876      openvswitch-ovn-host >= 0:2.6
> 00:00:31.877      openvswitch >= 0:2.6
> 00:00:31.877 package:
> vdsm-gluster-4.18.999-1162.gite9544ovirt-provider-ovn2e.el7.centos.noarch from
> check-custom-el7
> 00:00:31.877   unresolved deps:
> 00:00:31.877      vdsm = 0:4.18.999-1162.gite95442e.el7.centos

All of these seem like repoclosure false warning.

After all, vdsm = 0:4.18.999-1162.gite95442e.el7.centos is the exact
version of vdsm that is in the repo, right?

can't see it in http://resources.ovirt.org/pub/ovirt-4.1-snapshot/rpm/el7/x86_64/ while I see it in http://resources.ovirt.org/pub/ovirt-4.1-snapshot/rpm/el7/ppc64le/
so looks like vdsm is building different version of the arch packages. This hosuldn't happen.
Please check vdsm builders / publishers. They should deliver same version for both arches or noarch packages will fail dependencies.

FYI,
IIRC VDSM ppc64le isn't deployed to experimental because it fails CI due to mixing noarch pkg built by both ppc64le and x86_64, until this issue will be resolved on VDSM side ( it was resolved by spec change and reverted )
or the ppc64le build-artifacts job should built only the ppc64le rpms and not the noarch rpms.

Another possible option which is more complex and requires a major change in the way we use repoman, is to keep more versions back or not using the 'only-missing' option so having a few versions of VDSM should solve it.
This doesn't affect snapshot repos AFAIK
 



--
Sandro Bonazzola
Better technology. Faster innovation. Powered by community collaboration.
See how it works at redhat.com

_______________________________________________
Devel mailing list
Devel@ovirt.org
http://lists.ovirt.org/mailman/listinfo/devel



--
Eyal Edri
Associate Manager
RHV DevOps
EMEA ENG Virtualization R&D
Red Hat Israel

phone: +972-9-7692018
irc: eedri (on #tlv #rhev-dev #rhev-integ)