[ovirt-users] 3.6: GlusterFS data domain failover?

[Nir Soffer]

On Fri, Nov 20, 2015 at 1:09 PM, Wee Sritippho <wee.s at forest.go.th> wrote:
> Hi,
>
> My POC environment have 2 hosts - host A and host B, both are CentOS7.
> Installed oVirt 3.6 self-hosted engine. I manually created a 2-brick
> GlusterFS volume using both hosts and added it to my datacenter.

You are using replica 2 volume - this is not supported.

> I tried shutting down host A. The hosted-engine restarted in host B
> within 3 minutes, which is very cool. However, the GlusterFS data
> domain, which I set both 'Use Host' and 'Path' to the host A, is down
> along with it.

With replica 2, the file system becomes readonly as soon as one brick
is down.

>
> Here comes my questions:
> 1. How can I enable failover GlusterFS data domain?

Use replica 3 volume.

> 2. How can I reverse back to the state before adding the data domain?

Get both bricks up, remove the storage domain.

If this is the only storage domain, maybe adding another brick,
converting ti to replica 3 will be enough.

> The data domain is super persistent - I can't edit or delete it. I put
> it to maintenance mode but still unable to detach or destroy it because
> it requires me to remove the datacenter first. I tried but can't remove
> the datacenter either.

Is this the only datacenter? You cannot remove it if the hosted engine
doamin is using it.

I recommend separating the data center used by the hosted engine
domain and data center used for your vms.

> 3. Why can't I add another GlusterFS data domain? When I choose
> 'GlusterFS' as my 'Storage Type' every text field become grayed-out.

Maybe the selected host is down? Can you createa any other storage
domain?

> 4. When host A restarted again, I notice that 'Use Host' was changed
> from host A to host B instead. Is this an expected behavior?

Use host probably select the first host in the list of available hosts.

Nir