[lxc-users] cgmanager: cgm_list_children for controller=systemd, cgroup_path=user failed: invalid request

[Serge Hallyn]

Quoting Smart Goldman (ytlec2014 at gmail.com):
> 2015-01-15 0:11 GMT+09:00 Serge Hallyn <serge.hallyn at ubuntu.com>:
> 
> > cgmanager is an lxc project, so no worries.
> 
> Okay. and thank you for swift response.
> 
> > A few things - first, running cgroup-bin alongside cgmanager is
> probably a bad idea.
> 
> Oh really? It is better to remove either one?
> Another following 2 errors were output before installing these 2 things.
> sshd[4603]: pam_systemd(sshd:session): Failed to create session: No such
> file or directory
> systemd-logind[2957]: Failed to create cgroup name=systemd:/user/0.user: No
> such file or directory

0.user?  This is logging in as root?

> After executing 'apt-get -y install cgroup-bin cgmanager-utils', above 2
> errors were removed.
> 
> > Second, on 14.04 after install cgmanager
> > you unfortunately need to 'sudo restart systemd-logind', then
> > log out and log back in (or just ssh localhost) to get a new
> > cgroup.  After that you should be able to create unprivileged containers.
> 
> Now I tried 'sudo restart systemd-logind' and systemd-logind restarted.
> After that I logged out, log back and execute 'reboot'.
> However error log of cgmanager is still written on /var/log/auth.log.

Create the file /etc/default/cgmanager containing the text:

cgmanager_opts="--debug"

Restart cgmanager (sudo stop cgmanager; sudo start cgmanager) and then
login, and look at /var/log/upstart/cgmanager.log

> This error message is written even if I log out before reboot is executed
> or I execute reboot via control panel which is outside of ubuntu system.
> 
> > The message may be ignorable - the real question is, when you
> > login, do you get a custom cgroup?  What does /proc/self/cgroup
> > show?
> 
> I am sorry. How can I know whether I got a custom cgroup?
> Here is the contents of my /proc/self/cgroup.
> 
> root at myhost:~# cat /proc/self/cgroup
> 4:name=systemd:/user/1000.user/c1.session
> 3:freezer,devices,name=container:/12042
> 2:cpuacct,cpu,cpuset,name=fairsched:/12042
> 1:blkio,name=beancounter:/12042
> root at myhost:~#
> 
> I appreciate your help and I apologize for troubling you.

The /12042 is weird.  It looks like something libcgroup may have
done for you.

You may be able to work around this by adding the line:

env cgm_extra_mounts="-m name=systemd,name=container,name=fairsched,name=beancounter"

to /etc/default/cgmanager

-serge