On Thu, Dec 8, 2016 at 2:58 PM, Yedidyah Bar David <didi@redhat.com> wrote:
On Wed, Dec 7, 2016 at 6:46 PM, Sven Achtelik <Sven.Achtelik@eps.aero> wrote:
> Hi All,
>
>
>
> I’m planning to install oVirt 4 with the HC setup from this blogpost and did
> a demo setup.
>
>
>
> http://www.ovirt.org/blog/2016/08/up-and-running-with-ovirt-4-0-and-gluster-storage/
>
>
>
> I’ve managed to install everything just fine and thought that enabling a
> storage network would move all the bricks to that designated network, but it
> didn’t do anything.

What do you mean by this? AFAIK just "adding a network" should not do anything
in itself. If you refer to the section "Storage Network" from that post, then
it describes how to add the existing, pre-configured storage network, to the
engine management, not how to "create" it.

> I made that happen by destroying and recreating the
> volumes with entering the ip addresses of the storage network through the
> cli. Would that be the right way to do this ?

I'd say the right way is probably to plan beforehand and create your
storage network ahead. Choose nics for that, connect them to a separate vlan,
give them IP addresses, give these addresses names (different from the names
given to addresses of other nics), and use these names when configuring
the gluster storage.

> If that’s the case how could I
> move the engine bricks over to that network to get it off my ovirtmgmt
> bridge ?  Or is there may be more prefered way to complete the HC-Setup ?

If you provided IP addresses of the wrong nics and now want to change that,
then I have no idea - no personal experience with gluster. Adding Sahina
for this.

Once you have created the bricks with one network, changing it to another is a bit hard.
With latest version of gluster (3.9)  there's a command reset-brick that allows you to do that [1] - you will need to do this one brick at a time as the brick being reset will need to be killed first.

The safest way is to plan ahead and create your gluster volume using the network you wish to use.

[1] - https://bugzilla.redhat.com/show_bug.cgi?id=1266876
 

I suggest, in such cases, to always use names, never IP addresses. It's
a bit more maintenance work (of the dns and/or /etc/hosts), but provides
much more flexibility - if you used names, you can (in principle, didn't
try that myself) take everything down, change the names to point at the
IP addresses of the correct nics, then start all up.

Best,
--
Didi