[pkg-lxc-devel] Bug#859220: Script /usr/share/lxcfs/lxc.mount.hook fails with mount command error 32 and causes lxc-start to fail

[Evgeni Golov]

Hi Tony,

and everybody complains systemd would start too quickly ;)

Thanks for testing! I'll see that we add the right bits upstream and
then try to update all LXCFS branches we currently have
(jessie-backports, stretch, sid).

Regards
Evgeni

On Wed, Jun 21, 2017 at 09:22:01PM -0500, Tony Thedford wrote:
> Hello Evgeni,
> 
> I just verified that adding the 'cgroupfs-mount' service to the
> Required-Start and Required-Stop dependencies fixes the issue. Please see
> below. FYI - This same problem exists in Debian 9 in addition to Debian 8..
> and the below change to the sysvinit service files fixes both versions.
> 
> /etc/init.d/lxcfs:
> 
> Change from:
> ### BEGIN INIT INFO
> # Short-Description:    FUSE filesystem for LXC
> # Description:          FUSE filesystem for LXC
> # Provides:             lxcfs
> # Required-Start:       $remote_fs
> # Required-Stop:        $remote_fs
> # Default-Start:        2 3 4 5
> # Default-Stop:         0 1 6
> ### END INIT INFO
> 
> Change to:
> ### BEGIN INIT INFO
> # Short-Description:    FUSE filesystem for LXC
> # Description:          FUSE filesystem for LXC
> # Provides:             lxcfs
> # Required-Start:       $remote_fs cgroupfs-mount
> # Required-Stop:        $remote_fs cgroupfs-mount
> # Default-Start:        2 3 4 5
> # Default-Stop:         0 1 6
> ### END INIT INFO
> 
> Then run 'update-rc.d lxcfs defaults' to make changes active.
> 
> ---
> Tony
> 
> 
> 
> -------- Forwarded Message --------
> Subject: 	Re: Fwd: Re: Bug#859220: Script /usr/share/lxcfs/lxc.mount.hook
> fails with mount command error 32 and causes lxc-start to fail
> Date: 	Wed, 21 Jun 2017 19:07:47 -0500
> From: 	Tony Thedford <tony at accesslab.com>
> To: 	Evgeni Golov <evgeni at debian.org>
> 
> 
> 
> Your suspicions were correct! :-)
> After restarting lxcfs the entries do show up...
> Please see below:
> 
> root at theblob:~# ls -al /var/lib/lxcfs/cgroup
> total 0
> drwxr-xr-x 2 root root 0 Jun 21 17:13 .
> drwxr-xr-x 2 root root 0 Jun 21 17:13 ..
> 
> root at theblob:~# service lxcfs restart
> Stopping FUSE filesystem for LXC: lxcfs.
> Starting FUSE filesystem for LXC: lxcfs.
> 
> root at theblob:~# ls -al /var/lib/lxcfs/cgroup
> total 0
> drwxr-xr-x 2 root root 0 Jun 21 17:14 .
> drwxr-xr-x 2 root root 0 Jun 21 17:14 ..
> drwxr-xr-x 2 root root 0 Jun 21 17:14 blkio
> drwxr-xr-x 2 root root 0 Jun 21 17:14 cpu
> drwxr-xr-x 2 root root 0 Jun 21 17:14 cpuacct
> drwxr-xr-x 2 root root 0 Jun 21 17:14 cpuset
> drwxr-xr-x 2 root root 0 Jun 21 17:14 devices
> drwxr-xr-x 2 root root 0 Jun 21 17:14 freezer
> drwxr-xr-x 2 root root 0 Jun 21 17:14 net_cls
> drwxr-xr-x 2 root root 0 Jun 21 17:14 net_prio
> drwxr-xr-x 2 root root 0 Jun 21 17:14 perf_event
> root at theblob:~#
> 
> 
> 
> On 06/21/2017 02:36 AM, Evgeni Golov wrote:
> > Hi,
> > 
> > On Wed, Jun 21, 2017 at 01:01:36AM -0500, Tony Thedford wrote:
> > > Below you requested a listing of the file contents of /proc/self/cgroups and
> > > that file did not exist.. however I just noticed that there is a file named
> > > /proc/self/cgroup (without the 's' suffix).. and it has the following
> > > contents:
> > Right, sorry for the typo!
> > ANd thanks for all the output. I'll see how I can (or cannot) reproduce
> > that later. lxcfs *should* mount all entries from /sys/fs/cgroup/ (which
> > you have) to /var/lib/lxcfs/cgroup (which does not happen for you).
> > Maybe it's also just a race condition where lxcfs is started before
> > cgroup-mount?
> > 
> > Can you try restarting lxcfs and reporting the contents of
> > /var/lib/lxcfs/cgroup?
> > 
> > Evgeni
> 
> 
>