← Back to team overview

canonical-ci-engineering team mailing list archive

Re: Otto / LXC problems

 

On 10/11/2013 03:05 PM, Stéphane Graber wrote:
Hmm, I just connected to dx-autopilot-intel and I can't seem to
reproduce the issue there. The container starts and so does the whole
user session.
That is because it needed latest dbus on the image or the base container. Upgrading it inside a running container was not enough because the container must be restarted afterwards.
Using ubuntu desktop 20131011 to reprovision the containers fixed it.


On Fri, Oct 11, 2013 at 09:50:20AM +0300, Timo Jyrinki wrote:
Hi all,

Thanks for all the updates and work! I've dist-ugpraded the LXC
container and host with last night's lxc + dbus updates, but I'm still
getting:

http://10.97.0.1:8080/job/autopilot-saucy-daily_release/label=autopilot-intel/2511/console

-Timo


On Thu, Oct 10, 2013 at 7:14 PM, Vincent Ladeuil <vila+ci@xxxxxxxxxxxxx> wrote:
With dmesg really attached and jibel added in CC.

Vincent Ladeuil <vila+ci@xxxxxxxxxxxxx> writes:

Stéphane Graber <stephane.graber@xxxxxxxxxxxxx> writes:
     >> On Thu, Oct 10, 2013 at 10:12:39AM +0300, Timo Jyrinki wrote:
     >>> Hi,
     >>>
     >>> Thanks to the efforts yesterday, an lxc problem was identified and
     >>> fixed (https://lists.ubuntu.com/archives/saucy-changes/2013-October/011959.html)
     >>> during the night my time.
     >>>
     >>> However, cyphermox reported to me that he updated the containers and
     >>> the 'check' (autopilot) jobs of cu2d continue to fail. So the dbus <->
     >>> apparmor <-> lxc triangle is not yet in complete harmony. He suggested
     >>> checking the /var/log/upstart/lightdm.log and sending it to stgraber /
     >>> security team. So here goes:
     >>>
     >>> http://pastebin.ubuntu.com/6216936/
     >>>
     >>> It's taken from dx-autopilot-intel and the latest container
     >>> saucy-i386-20131010-0216.
     >>>
     >>> -Timo
     >>>
     >>> ps. thanks Evan for the ubuntu-engineering post

     >> Can you make sure the host is up to date,

     > It is.

     >> reboot it after that

     > Done.

     >> (since the dbus apparmor changes came through a kernel change) and
     >> then tell me what version of LXC is on the host

     > root@dx-autopilot-intel:/var/lib/jenkins# apt-cache policy lxc
     > lxc:
     >   Installed: 1.0.0~alpha1-0ubuntu10
     >   Candidate: 1.0.0~alpha1-0ubuntu10
     >   Version table:
     >  *** 1.0.0~alpha1-0ubuntu10 0
     >         500 http://us.archive.ubuntu.com/ubuntu/ saucy/main i386 Packages
     >         100 /var/lib/dpkg/status

     >> and attach the "dmesg" output right after a failure.

     > dmesg attached.

     >> If it's really still an apparmor profile issue, there will be denials in
     >> the dmesg output, if not, then it's something else that's breaking your
     >> tests.

     > Now, otto (which is creating the lxc container) mounts an iso and as
     > such requires a loop device. To get that it disables:

     > root@dx-autopilot-intel:/etc/apparmor.d/disable# ls -lart
     > total 8
     > lrwxrwxrwx 1 root root   33 Jun  4 10:34 usr.sbin.rsyslogd -> /etc/apparmor.d/usr.sbin.rsyslogd
     > drwxr-xr-x 9 root root 4096 Oct 10 13:14 ..
     > lrwxrwxrwx 1 root root   33 Oct 10 13:15 usr.bin.lxc-start -> /etc/apparmor.d/usr.bin.lxc-start
     > drwxr-xr-x 2 root root 4096 Oct 10 13:15 .

     > Not sure how that interacts with the profiles but without them the
     > container can't be started and use its iso.

     >           Vincent



--
Jean-Baptiste
IRC: jibel


References