On 04/05/2014 02:33 PM, Sven Kieske wrote:
The cpu model actually gets read from
a libvirt xml file.
Extensive documentation about this can be found here:
http://wiki.libvirt.org/page/Libvirt_identifies_host_processor_as_a_diffe...
iirc, it is still a pass-through to qemu-kvm, which accepts -cpu
<cpu_model>.
HTH
Am 04.04.2014 17:32, schrieb Itamar Heim:
> On 03/31/2014 06:57 PM, Robert Story wrote:
>> On Wed, 19 Mar 2014 10:13:28 -0400 (EDT) Omer wrote:
>> OF> cpu name is cluster level, and usually the lowest common
>> denominator of
>> OF> the hosts in cluster, to make sure migration works in the cluster.
>>
>> Is there and documentation on the relationship between cpu types and
>> compatibility? Right now I have a cluster per cpu type, and it might make
>> sense to merge some of them if the performance hit was minimal.
>
> that would be a question for qemu-kvm - they provide the cpu models we
> base this on.
> you can check which cpu flags you get exposed in the guest of the higher
> cpu model, then check if you think your applications would care about
> them...