[systemd-devel] [RFC] [PATCH] cgroup: don't trim cgroup trees created by someone else

[Lennart Poettering]

On Wed, 05.11.14 12:27, Lennart Poettering (mzerqung at 0pointer.de) wrote:

> > it doesn't seem to be correct solution either. systemd will happily remove cgroup
> > in which there are processes.
> 
> Oh. right, systemd is stricter there than I remembered: we will
> actually migrate the PIDs before removing the cgroup.
> 
> I figure we need to figure out a way how we can make a cgroup capable
> for embedding their own systemd instances, so that the controller
> memberships cover all hierarchies.
> 
> I need to think about this.

OK, after some thinking and some discussions and more thinking, I now
added this:

http://cgit.freedesktop.org/systemd/systemd/commit/?id=a931ad47a8623163a29d898224d8a8c1177ffdaf

With this in place libvirt-lxc should just work, if it properly
creates the root cgroup of is containers with machined's
CreateMachine() call.

Lennart

-- 
Lennart Poettering, Red Hat