On Wed, Apr 16, 2014 at 11:24:22AM +0200, Sandro Bonazzola wrote:
Il 03/04/2014 21:20, Gianluca Cecchi ha scritto:
> Hello,
> I know that AIO setup is only for experimenting and trying but I find
> it very useful in several situations. So having a 3.3.4 All-In_one
> installation on fedora19 and wanting to update it to 3.4 I have some
> doubts.
> In release notes I see for general updates:
>
> On Fedora 19 you'll need to enable fedora-updates repository for
> having updated openstack packages --> OK. It is normally already
> enabled
>
> On Fedora 19, you'll need to enable fedora-virt-preview repository for
> using Fedora 19 as node on 3.4 clusters --> how do I managed this?
> This is indeed a node too, so I have to enable virt-preview for it?
> Correct?
Correct
>
> BTW: Is this note still true in general or did the related virt
> packages go into fedora-updates, or is it planned in coming
> days/weeks?
It's still true in general. libvirt people said they won't update F19 with the
virt-preview packages.
> When engine-setup is run and detects that a newer version is available
> and outputs that I have to run
>
> yum update ovirt-engine-setup
>
> is it correct to say that in AIO setups I should actually run
>
> yum update ovirt-engine-setup ovirt-engine-setup-plugin-allinone
On all-in-one think that it may be useful a general "yum update" in order to
get libvirt, vdsm and ovirt-engine-setup*
If you want to update only the minimum for the upgrade, better go with yum update
"ovirt-engine-setup*"
>
> or is it the latest package not crucial to be updated before new
> engine-setup is run?
>
> At the end of engine update can I simply put the AIO server into
> maintenance (as a node) and run "yum update" so that vdsm (aka node)
> packages are updated too and so the server will become a 3.4 enabled
> node after reboot?
I think you should move the system in maintenance before running engine-setup.
I also don't think that a full reboot is needed, I think that just service vdsmd
restart should be enough.
But Dan can tell you if the reboot is needed.
Actually, vdsm restarts itself (and supervdsm) on rpm upgrade, so if
anything explicit is required (certainly a reboot), it's a bug.