[libvirt] cpuset / numa and qemu in TCG mode

[Daniel P. Berrange]

On Tue, May 12, 2015 at 05:27:34PM +1000, Tony Breeds wrote:
> On Mon, May 11, 2015 at 01:14:58PM +0200, Martin Kletzander wrote:
> 
> > Determining this by version might not be reliable, but more
> > importantly working around bug in underlying software is something
> > that shouldn't be done at all IMHO.  Let the maintainers backport
> > whatever needs to be done.
> 
> I agree with you in an ideal world but there are times when we do need
> to add work arounds in $project_x to work around issues in $project_y.
>  
> > >Ther nova side will be pretty easy regardless.
> > >
> > >I'd say the best solution would be to back port the 'fix' but that seems like a
> > >lot of effort given the number of distros and libvirt versions potentiall
> > >involved.
> > >
> > 
> > If you want the fix to be distro-agnostic, there's nothing easier than
> > back-porting the fix into our upstream maintenance branches.  Those
> > should make the life of distro maintainers easy (although it looks
> > like not many distros use it).
> 
> And this is part of the problem.  If I understand correctly Ubuntu cloud-archive
> is using libvirt 1.2.12 which is *NOT* a maintenance release so that leaves us
> with doing an additional backport to 1.2.12 and getting the cloud-archive team
> to take it[1]  or Adding a hack to nova.  And that's just Ubuntu It's hard to
> say for sure that some vendor isn't running libvirt 1.2.12 also.
>  
> > Having said that I'm not sure which commit(s) are those that need to
> > be back-ported.  Having known your libvirt version, it shouldn't be
> > too hard looking for the differences and finding the right commit.
> > When back-porting request is made on the list, it is usually acted
> > upon.  If you can't find the exact commit, let me know and I'll do my
> > best to help.
> 
> So a git bisect points at:
> ---
> commit a103bb105c0c189c3973311ff1826972b5bc6ad6
> Author: Daniel P. Berrange <berrange at redhat.com>
> Date:   Tue Feb 10 15:59:57 2015 +0000
> 
>     qemu: fix setting of VM CPU affinity with TCG
> ---
> 
> A small amount of reading implies to me that we'd be looking at backporting
> a103bb105c0c189c3973311ff1826972b5bc6ad6 to any maintenance branch that contains
> b07f3d821dfb11a118ee75ea275fd6ab737d9500.  Which I think is 1.2.13 only, but I
> could be wrong.
> 
> If you don't beat me to it I'll request that backport to 1.2.13 *and* ask the
> Ubuntu guys to take it as well.
> 
> I have to admit I'm still in 2 minds on the nova side.  Adding a wart to the
> libvirt driver for this specific bug for distros/vendors that are using 1.2.12
> seems a bit gross but ....

I think this is something distros should really fix in their libvirt
packages - backporting this kind of fix is exactly the kind job of
maintainers should be doing.

Particularly since it only affects TCG mode, I don't think Nova really
should try to workaround it

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 :|