2010/3/17 Daniel P. Berrange <berrange(a)redhat.com>:
On Wed, Mar 17, 2010 at 12:02:27PM +0100, Matthias Bolte wrote:
> 2010/3/17 Daniel P. Berrange <berrange(a)redhat.com>:
> > On Wed, Mar 17, 2010 at 11:05:35AM +0100, Matthias Bolte wrote:
> >
> > Conceptually this doesn't make sense. The max downtime parameter is a
> > live tunable that a management app will set based on the conditions at
> > the time of migration. Suitable value is dependant on the workload of
> > the guest (how quickly it is dirtying RAM) and network bandwidth between
> > hosts (how quickly RAM is transferred). You would only increase the
> > max downtime if the VM was not able to complete migration promptly
> > enough. So it shouldn't be persistent, and I don't think there's
any
> > compelling need for a getter either.
> >
> > Regards,
> > Daniel
> >
>
> True. I just had the API POV in mind and forgot about the rest.
>
> So we add a new function to configure a per-migration option because
> we cannot change the existing migrate functions to add a new
> parameter.
It isn't sufficient to set this in virDomainMigrate, since it needs to
be tuned (possibly multiple times) while migration is taking place.
Regards,
Daniel
Ah, now I understand how this is supposed to be used and how to it should work!
With that concept in mind the original patch is fine.
Matthias