[Openstack] Nova guest NUMA placement questions

Daniel P. Berrange berrange at redhat.com
Fri Sep 26 14:04:59 UTC 2014


On Fri, Sep 26, 2014 at 01:43:26PM +0000, Robert van Leeuwen wrote:
> > I'm new to OpenStack development and would like to test the upcoming
> > NUMA features in Nova, particularly with regards to this blueprint:
> >
> > ....
> >
> >4. How to test whether a guest topology is actually using the memory
> > from the NUMA node it is specified to run on?
> >    - Can this be verified at the libvirt level?
> >   - Is there a way by querying /proc/PID the memory map/cpu allocation?
> 
> You can run numastat, e.g. show memory allocation of all qemu processes:
> numastat -p qemu
> 
> Note that there is a numad daemon that actually also optimizes a system
> for numad.
> Seems to be pretty efficient at placing machines within our cloud (looking
> at numastat)

You should not run numad on any Nova Juno compute host where you are making
use of Nova's NUMA capabilities. They will fight over placement leading
to pain & suffering because numad will change things behind nova's back
leading the nova schedular to make bad decisions for guest placement.

The NUMA work is intended to obsolete the need to use numad, since it has
a world view of the entire cloud and so does intelligent picking of the
hosts before even starting the guest.

Regards,
Daniel
-- 
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|




More information about the Openstack mailing list