[Pkg-libvirt-maintainers] Bug#732666: Bug#732666: severity:serious, since it breaks the upgrade

[phep]

Hi,

Le 01/01/2014 20:24, Guido Günther a écrit :
> If you don't have memory cgroups enabled in the kernel you already broke
> the daemon restart yourself by configuring it in. The discussion is moot
> since we'll detect this dynamically in the future (with the next upload)
> but it's rather a configuration error than a bug.

The pending patches should avoid the problem Solveig and I met but there 
might be one new problem hiddent : before the upgrade we had a unique mount 
point for all cgroups and (as far as I can get it) the "memory" cgroup could 
be added at runtime, as needed for the configuration of some guest, without 
any further modification. Now, to have /sys/fs/cgroup/memory, one _has_ to 
have cgroup_enabled=memory on the kernel command line.

For example, I have the following in one lxc guest config file (managed with 
the standard lxc tools, not virsh) :
...
lxc.cgroup.memory.limit_in_bytes       = 256M
lxc.cgroup.memory.memsw.limit_in_bytes = 1G
...

I'm not sure this would still work without the cgroup_enabled=memory option 
set. (I have not found time to test it yet, I'm on a trip and avoid 
rebooting the laptop without major need - more time next week-end for this).

Yet, as already acknowledged before, I'm anything but a kernel wizard so 
that I may miss something obvious or be utterly in the blue.

Cheers,

phep