openstack team mailing list archive

[Lloyd Dewolf <lloydostack@xxxxxxxxx>]

On Tue, Apr 30, 2013 at 2:14 PM, Dugger, Donald D
<donald.d.dugger@xxxxxxxxx> wrote:
> I don’t claim to be an expert on OpenStack on OpenStack but I don’t believe
> you need VTx at all.  IPMI, yes you need that, but not VTx.

How would (physical) IPMI come in to this equation? If I understand
here we are talking about using virtualization on a single server to
gain experience with configuring and operating OpenStack clusters.


On Tue, Apr 30, 2013 at 1:12 PM, Chris Bartels
<chris@xxxxxxxxxxxxxxxxxxxxxx> wrote:
>
> tl;dr-
>
> 1.       Does the nested KVM running inside the OpenStack that runs inside
> another OpenStack get to take advantage of the VT-x of the host CPU?

AFAIR, there has been nested guest support in kvm_amd for a while, and
more recently kvm_intel also has support...
These blog posts from a year ago seem relevant:
http://kashyapc.wordpress.com/2012/01/14/nested-virtualization-with-kvm-intel/
http://kashyapc.wordpress.com/2012/01/18/nested-virtualization-with-kvm-and-amd/



> 2.       Does OpenStack on OpenStack running KVM in each need VT-x on the
> host CPU at all to run properly?

Sort of, hardware support is required to truly use KVM, but when we
say KVM we're lumping in QEMU. KVM fails back to QEMU.

AFAIK, other virtualization technologies like vmware and virtualbox
require VT-x/amd-v for 64-bit guest virtualization, and I think some
Windows guests require tech like Extended Page Table (EPT). I've
previously played with https://github.com/lorin/openstack-ansible


If it was me, I would be concerned that without hardware
virtualization the performance would be so poor that OpenStack on
OpenStack would be too frustrating to use -- though you might find
some interesting bugs related to race conditions and timeouts :p


Hope that helps,
--
@lloyddewolf
http://www.pistoncloud.com/