[Users] virt-manager migration

[Ryan Barry]

On 04/03/2014 03:58 PM, Jeremiah Jahn wrote:
> Well, that was fun. So I let the ovirt engine install to a running
> host that already had kvm/libvirt running on it. Don't ask why, but it
> did happen.  After figuring out how to setup a sasl user/password and
> adding qemu to the disk group I could startup all of my guests again.
>    My host now shows up in the list of hosts, but has "One of the
> Logical Networks defined for this Cluster is Unreachable by the Host."
>   error sitting on it.  ovirt-node-setup also tells me I should setup a
> network.   I currently have 6 bridges running on this thing all one
> for each vlan. I'm unsure as to how to meld the 'bondX' in
> ovirt-node-setup with my current network configuration to resolve the
> error.  esp given that I don't actually want to bond any of my NIC's
> together at this point.  I do realise I'm doing this the hard way.  My
> goal at the moment is to just get the host to fully report in the
> engine, at which point I think I'll be able to use v2v to finish up
> the rest.
>
> Thanks for any suggestions or pointers.
>
That's ok. I probably would have skipped ovirt-node-setup (which is 
intended to run on the oVirt Node ISO) and used the "New" wizard from 
the Engine to add it (which would install requisite RPMs, etc).

The "one of the networks is not reachable" error isn't really my area, 
but it's probably looking for the host to be reachable by IP on a bridge 
called ovirtmgmt.

An engine developer can verify this, but I'd guess that adding a virtual 
NIC to whatever VLAN has the IP the engine sees and bridging that to 
ovirtmgmt with the appropriate address would work. It may only need the 
right bridge (ovirtmgmt, probably) defined. I've never experimented with 
this aspect of it, to be honest.
>
> On Thu, Apr 3, 2014 at 10:29 AM, Ryan Barry <rbarry at redhat.com> wrote:
>>> From: "Jeremiah Jahn" <jeremiah at goodinassociates.com>
>>> To: users at ovirt.org
>>> Sent: Wednesday, April 2, 2014 8:38:02 PM
>>> Subject: [Users] virt-manager migration
>>
>>>
>>
>>> Anyway, long story short. I'm having a difficult time finding
>>> documentation
>>> on migrating from virt-manager to oVirt. As well as installing ovirt-nodes
>>> manually. I'd love to find this perfect world where I can just install the
>>> ovirt-node RPMs on my already running Hosts and begin to have them managed
>>> by the oVirt engine. Without and serious downtime.
>>
>> The usual way is to go through virt-v2v. Essentially, you'd install the
>> engine somewhere and configure a storage domain (the properties of which
>> vary, but it's UUIDed and the UUID must match the engine) to bring the
>> datacenter up, then add an export domain (which is also UUIDed).
>>
>> Once an export domain is created, virt-v2v can move your VMs over, but with
>> downtime.
>>
>> As far as turning your existing hosts into nodes, adding them from the
>> engine is the easiest way (there's a wizard for this). It's possible to
>> install the ovirt-node RPMs directly, but they take over your system a bit,
>> and it's probably not what you're looking for. The engine can manage regular
>> EL6/fedora hosts.
>>
>> But registering to the engine will reconfigure libvirt, so the general path
>> is:
>>
>> Install engine.
>> Live-migrate VMs off one of your hosts.
>> Add that host as a node.
>> virt-v2v machines than can take downtime (can you get a maintenance window)?
>> Bring them up on the new node.
>> Repeat until your environment is converted.