On 19 Dec 2014, at 00:09, Itamar Heim <iheim(a)redhat.com>
wrote:
> On 12/18/2014 05:13 PM, Markus Stockhausen wrote:
> Hello,
>
> we just build a new cluster with FC20 + virt-preview repos enabled.
> Idea behind that is to enable snapshot live merge feature. This
> seems to work quite well.
>
> The only culprit is the Windows activation. For some reasons the
> VM hardware of the old qemu 1.6/seabios 1.7.3 hypervisors is
> different to the new qemu 2.1/seabios 1.7.5. So the OS is no longer
> activated.
>
> We already activated the VMs twice. During first install on VMware
> and then afterwards after the migration to OVirt. I have no problem
> to reactivate them a third time. I have more fears that I must
> reactivate them with each new hypervisor generation. That can
> be quite a lot of phone calls.
>
> An interesting article from the Proxmox guys can be read here:
>
http://forum.proxmox.com/archive/index.php/t-19743.html
> Conclusion of the discussion: After you go from qemu 1.7 to 2.1
> you can force the old hardware layout using -M pc-i440fx-1.7
>
> Looking at the qemu command line in OVirt I can see our VMs
> are fired up with pc-1.0,accel=kvm,usb=off - regardless of an
> old (FC20) or a new hypervisor (FC20+virt-preview). So I would
> guess that everything should work without a new reactivation.
>
> Any ideas?
>
> Markus
>
>
>
>
>
>
> _______________________________________________
> Users mailing list
> Users(a)ovirt.org
>
http://lists.ovirt.org/mailman/listinfo/users
for .el6/.el7 we use the stable -m machines (-m rhel6.5.0, -m rhel-6.6.0, etc.)
fedora uses "pc" indeed. you can change it via config, but I don't know
fedora is keeping these stable for the other definitions / how to map them to cluster
levels
Change ClusterEmulatedMachines via engine-config to include the type suggested above
instead of pc. It would be more close to the hw we still use in 3.5 on el6/7