----- Original Message -----
> From: "Digimer" <[email protected]>
> To: "General Linux-HA mailing list" <[email protected]>
> Cc: "David Vossel" <[email protected]>
> Sent: Thursday, May 23, 2013 4:45:10 PM
> Subject: Re: [Linux-HA] vm live migration without shared storage
>
> On 05/23/2013 03:00 PM, David Vossel wrote:
> > Hey,
> >
> > I've been testing libvirt live migration without shared storage in Fedora
> > 19 alpha. Specifically this feature,
> > https://fedoraproject.org/wiki/Features/Virt_Storage_Migration. They've
> > done a lot of work to make it a much more solid option than it has been in
> > the past. I had to work through a few bugs with the guys over there, but
> > it's working great for me now.
> >
> > So, do we want to use this in HA? It would be trivial to add this to the
> > VirtualDomain resource agent, but does it make sense to do this?
> > Migration time, depending on network speed and hardware, is much longer
> > than the shared storage option (minutes vs. seconds). I don't mind
> > adding this support to the agent, but I wanted to get peoples' feedback
> > and make sure this is something we want before making that effort.
> >
> > -- Vossel
>
> Without shared storage, I personally don't see much of a point. This
> strikes me as a "move" tool, more than a "migrate" tool.
Well, it's kind of equally both. It's a "move the vm while it is running to
another node, flip a switch after the move is complete to live migrate the vm
to the other node." sort of thing. If it was just move without the instant
live migration this would be no different than taking a snapshot of the vm,
pausing it, scp the image and state to another box, and starting it again.
This is a true live migration though.
> It's value is
> certainly very high, and I could see it being an excellent way to move a
> VM from one cluster (and it's shared storage) to another.
>
> That said, without some way to keep the VM in sync,
> I can see very few
> use cases that would make this useful to HA.
I would think vm disk IO would be improved if the image didn't live on shared
storage. Keeping the image in sync across the cluster might be tricky. We
would undefine the vm on the source after the migration, but if things got
hairy with node failures there is the potential an older image could boot.
-- Vossel
>
> digimer
>
> --
> Digimer
> Papers and Projects: https://alteeve.ca/w/
> What if the cure for cancer is trapped in the mind of a person without
> access to education?
>
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
