Created attachment 827733 [details]
output of 'yum repolist'
Description of problem:
Since the oVirt 3.3.1 release and the RHEL repo update on 11/21, users running RHEL 6 on their KVM Host are unable to execute a package update, due to dependency conflicts.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Install RH 6.4 on host, prior to Nov 21
2. Add host to oVirt 3.3.0 engine
3. Execute yum update on host
See the attached 3 files, which include a repolist, output from "yum update", and output from "yum update --skip-broken". Note in particular that the update is trying to update glusterfs-cli, but there is no version to match the latest glusterfs provided by the current RHEL 6 repository. vdsm requires glusterfs-cli.
Created attachment 827734 [details]
Output of 'yum update'
Created attachment 827735 [details]
Output of 'yum update --skip-broken'
ovirt should not ship vdsm packages for i686. They only create confusion.
According to the logs
Protected multilib versions: vdsm-python-4.12.1-4.el6.i686 != vdsm-python-4.13.0-11.el6.x86_64
an i686 blocks the requested update of the x86_64 one.
Similarly, and installed glusterfs-libs-3.2.7-1.el6.i686 from epel blocks the update of glusterfs-libs.
Error: Package: glusterfs-cli-3.4.0-8.el6.x86_64 (@glusterfs-epel)
Requires: glusterfs-libs = 3.4.0-8.el6
Removing: glusterfs-3.4.0-8.el6.x86_64 (@glusterfs-epel)
glusterfs-libs = 3.4.0-8.el6
Updated By: glusterfs-18.104.22.168rhs-1.el6.x86_64 (rhel-x86_64-server-6)
Removing: glusterfs-libs-3.4.0-8.el6.x86_64 (@glusterfs-epel)
glusterfs-libs = 3.4.0-8.el6
Updated By: glusterfs-libs-22.214.171.124rhs-1.el6.x86_64 (rhel-x86_64-server-6)
glusterfs-libs = 126.96.36.199rhs-1.el6
Available: glusterfs-3.2.7-1.el6.i686 (epel)
glusterfs-libs = 3.2.7-1.el6
Bottom line, it seems a yum configuration issue more than a vdsm dependency bug. Please verify that, Bob.
Created attachment 827895 [details]
Output of "rpm -qa"
I have no vdsm*.i686 packages on my Host system. I have attached output of "rpm -qa".
At Vijay Bellur's suggestion, I edited the glusterfs-epel repo to point to release 3.4.1 instead of 3.4.0 (e.g. http://download.gluster.org/pub/gluster/glusterfs/3.4/3.4.1/EPEL.repo/epel-$releasever/$basearch/), and after that yum was able to update cleanly, and things appear to be working (of course I am not using GlusterFS).
I still have questions, however:
1. Why/how did glusterfs-cli disappear from the 3.4.0 repository? If due to refactoring, are they missing a Provides in their RPMs? Or is it oversight?
2. Why is glusterfs-cli in the 3.4.1 repository, while not in 3.4.0?
3. How can we address this in oVirt, in order to avoid admins having to edit their yum config files on every Host?
It has been suggested that #3 can be addressed by the glusterfs team providing a "stable" repository pointer that can be consumed by the oVirt Host configuration. This would avoid similar issues in future (assuming the glusterfs team's plan is to continue to branch repositories for every minor release).
Although this problem showed up first in the RHEL repositories, as part of the upgrade to RHEL 6.5, it is likely to show up for CentOS hosts as well, as soon as CentOS produces 6.5. So it's best for us to try and address this within oVirt quickly so that the pain to CentOS users can be minimized.
> 1. Why/how did glusterfs-cli disappear from the 3.4.0 repository? If due to
> refactoring, are they missing a Provides in their RPMs? Or is it oversight?
If you're referring to gluster RPMs that were in the EPEL repository, those were withdrawn in concert with the release of RHEL 6.5, which has RHS client-side RPMs. RHS client-side does not include the glusterfs-cli RPMs. N.B. that Fedora/EPEL policy does not allow shipping packages in EPEL that are in RHEL.
> 2. Why is glusterfs-cli in the 3.4.1 repository, while not in 3.4.0?
The glusterfs-cli RPM has not disappeared from the 3.4.0 repository on download.gluster.org.
> 3. How can we address this in oVirt, in order to avoid admins having to edit > their yum config files on every Host?
CentOS is doing their own packaging of community GlusterFS — clones of the download.gluster.org packages now that GlusterFS packages are no longer in EPEL. Their GlusterFS packages will supplant the RHS packages in RHEL 6.5.
Independent of that, the sample repo file that's at http://download.gluster.org/pub/gluster/glusterfs/LATEST/EPEL.repo has baseurl lines in the form of http://download.gluster.org/pub/gluster/glusterfs/LATEST/EPEL.repo/.... which, if I'm not mistaken, should simplify things for most people. (I hope all our releases going forward will be backwards compatible.)
I have determined that the problem creeps in here:
I just followed the Quick Start Guide: http://www.ovirt.org/Quick_Start_Guide#Install_oVirt_Node
That involves installing an RPM on the host (the Quick Start lists the Fedora variant, so I adapted to use the EL RPM). It contains:
% rpm -qlp http://ovirt.org/releases/ovirt-release-el6-8-1.noarch.rpm
So there's your culprit. At the moment the delivered glusterfs-epel.repo configures the host/node explicitly to use 3.4.0. One solution is to change it to use LATEST.
so how can I help here? glusterfs is outside of my natural habitat.
Oh, sorry. Who owns ovirt-release.rpm? I like the idea of changing it to point to a newer glusterfs-epel (either 3.4.1 for stability or LATEST for easy updates).
I was able to update by doing this:
# yum update ovirt\*
this one updates ovirt-release to ovirt-release-el6-9-1
# yum clean all
# yum update glusterfs\*
# yum update
(In reply to Kapetanakis Giannis from comment #11)
Marking as verified
oVirt 3.3.2 has been released resolving the problem described in this bug report.