Re: Live migration using a specified networking adapter

["Daniel P. Berrangé" <berrange@redhat.com>]

On Mon, Apr 19, 2021 at 08:06:23PM +0100, Dr. David Alan Gilbert wrote:
> * Jing-Wei Su (jwsu1986@gmail.com) wrote:
> > Hello experts,
> > 
> > 
> > I have a network topology like this diagram.
> > 
> > When start live migration moving a VM from Host A to B,
> > 
> > the migration process uses either 10GbE (10.0.0.1) or 1 GbE (10.0.0.2),
> > 
> > but the user cannot specify the source NIC by current migrate command.
> > 
> > 
> > To solve the problem, my rough idea is to add a source ipv4:port argument,
> > 
> > the migration command seems like
> > 
> > ```
> > 
> > migrate -b tcp:10.0.0.1:4444 -d tcp:10.0.0.3:4444.
> 
> I'm not sure what the OS lets us do, but if it lets us pick the IP and
> port I think that would work; I don't think you need another tcp:
> since we already know which protocol we're using.
> 
> > 
> > ```
> > 
> > Is it an available solution? Or, is there any concern and sugesstion?
> > 
> > Besides the idea, is there any good way to this issue?
> 
> It's unusual; I don't think I've seen anyone ask for it before.
> I assume there's a wayto get the host network stack to prefer
> the 10GbE interface.
> Or to use separate subnets; rememember that each interface
> can have multiple IP addresses.

Yes, the recommended way is to have separate subnets for the
two NICs, and then use the IP address associated with the
fast NIC. Network routing will ensure migration traffic flows
over the main NIC. This is known to work well as OpenStack
uses this approach with QEMU/libvirt already.

Regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|