Open Stack

On Fri, Sep 07, 2012 at 11:25:45AM +0200, Thierry Carrez wrote:
> Chuck Short wrote:
> > On 12-09-06 02:08 PM, David Kang wrote:
> >>   Currently cgroups_path is hard coded in virt/disk/api.py as
> >> "/sys/fs/cgroup/".
> >> In REHL 6.x, the default path is "/cgroup/".
> >> How about making it configurable in nova.conf file?
> 
> Note that we should avoid introducing new configuration options at this
> point in the cycle, so that documentation has a chance to be correct at
> release time...
> 
> Maybe for Folsom that should be patched in the distros that require it,
> and made configurable in Grizzly ?

Well all that LXC hotplug code can really just be ripped out. Bind
mounting a device into the container's /dev as seen from the host
OS really cannot work. The container is running in a separate
filesystem namespace with a private /dev mount on tmpfs. So whatever
you do to the /path/to/container/filesystem/dev on the host will
never be visible to the container.

Once the kernel supports attaching to a container namespace using
/proc/self/ns/mount, then libvirt will officially support LXC hotplug,
at which point libvirt can take care of cgroups too.

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 :|