[Users] Update from 3.2 to 3.3 (CentOS 6.4)

[H. Haven Liu]

Thanks for the discussion. But for those of us that are not using gluster, but just good ol' NFS, is updating simply "yum update ovirt-*"?

On Sep 16, 2013, at 2:54 PM, Mike Burns <mburns at redhat.com> wrote:

> On 09/16/2013 05:26 PM, Joop wrote:
>> Mike Burns wrote:
>>> On 09/16/2013 04:30 PM, Joop wrote:
>>>> H. Haven Liu wrote:
>>>>> Hello,
>>>>> 
>>>>> Is there any recommended procedure for updating from 3.2 to 3.3
>>>>> (namely on a CentOS 6.4 system), or anything one should be careful
>>>>> when doing such update?
>>>> Be careful when you use glusterfs, not nfs over glusterfs, but read the
>>>> release-notes. It should have something to say about glusterfs domain
>>>> not (yet) working on el6.
>>>> Saw a small discussion on irc just yet and my two cents are that you
>>>> can't add el6 support to ovirt in release 3.2 and then withdraw it with
>>>> 3.3 and say well just wait for Centos/Rhel-6.5.
>>> 
>>> We haven't removed any functionality in 3.3.  In 3.2, we added support
>>> for gluster domains through a POSIXFS interface.  In 3.3, we're adding
>>> a feature where we support gluster natively.  This works in Fedora,
>>> but is not available on EL6.  The POSIXFS option still exists
>> You're right but what about users who want to use the (much) improved
>> speed of the gluster domain over the POSIXFS interface? They are left
>> out or they should move to Fed19 which I would do but I need to convince
>> a couple of other people as well and they aren't going to agree.
> 
> I understand the complaint, honestly.  And we're working on a solution so that it will work for people on EL6.
> 
> From the perspective of whether we should release with this limitation or not, I'd point out that by not releasing, we'd be preventing everyone from using any of the new features until we get a solution for this. I'd rather release and make it available for everyone now and say that the Gluster domain for EL6 will come as soon as we can work out the dependency issues.
> 
>> 
>>> 
>>>> The kernel can come from elrepo so that is not a burden for the ovirt
>>>> team, qemu/libvirt should be build by the ovirt team and be available
>>>> from the ovirt repo. At the moment I also see/saw Jboss-7.1.1
>>>> qemu/libvirt wouldn't be the first packages to be in the ovirt-repo
>>>> which are also in the main distributions repos.
>>> 
>>> We're trying to work out a way to do this in a consistent manner going
>>> forward.  We should have a solution soon, but in the meantime, the
>>> other functionality and features should work on both Fedora and EL6.
>> Thanks for the clarification and I'm waiting eagerly for what/when the
>> solution comes out.
> 
> There have been a few considerations for solving this including rebuilding pure upstream or fedora packages for EL6.  That is a risky solution in my mind since there are rather large deltas between Fedora and EL6.  We're looking at whether we can have a "virt-preview" type of repo for EL6 similar to what exists today for Fedora[1].
> 
> 
> Thanks
> 
> Mike
> 
> [1] http://fedoraproject.org/wiki/Virtualization_Preview_Repository
>> 
>> Joop
>> 
>> _______________________________________________
>> Users mailing list
>> Users at ovirt.org
>> http://lists.ovirt.org/mailman/listinfo/users
> 
> _______________________________________________
> Users mailing list
> Users at ovirt.org
> http://lists.ovirt.org/mailman/listinfo/users