libvirt users - Jul 2012 - Problem halting/restaring a lxc container from within

Hi,

I've been making some tests with libvirt and LXC and found some problems
when halting/restarting a LXC container from within.

Basically, on a Ubuntu 12.04 system with libvirt installed as package
(0.9.8), I've created a basic container image with:

 lxc-create -t ubuntu -n lxc

And started it using the libvirt XML listed below and the following command:

 virsh -c lxc:// create lxc.xml

I can access the container console normally but when I issue a halt or
reboot command from inside, the container initiates the halt/reboot
execution but it does not finishes it properly.

For example:

$ sudo halt -p
[sudo] password for ubuntu:

Broadcast message from ubuntu at my-container
	(/dev/pts/0) at 23:19 ...

The system is going down for power off NOW!
ubuntu at my-container:~$ acpid: exiting
 * Asking all remaining processes to terminate...    [ OK ]
 * All processes ended within 1 seconds....          [ OK ]
 * Deconfiguring network interfaces...               [ OK ]
 * Deactivating swap...                              [fail]
 * Unmounting weak filesystems...                    [ OK ]
umount: /run/lock: not mounted
mount: / is busy
 * Will now halt

The container seems to finish the shutdown process but libvirt does not
seem to be signaled about it (virsh shows the container is still
executing). Something similar happens with reboot.

However, if a start that container with:

 lxc-start -n lxc

and do the same test, it works perfectly:

# sudo halt -p
[sudo] password for ubuntu:

Broadcast message from ubuntu at my-container
	(/dev/lxc/console) at 23:17 ...

The system is going down for power off NOW!
ubuntu at my-container:~$ acpid: exiting
 * Asking all remaining processes to terminate...        [ OK ]
 * All processes ended within 1 seconds....              [ OK ]
 * Deconfiguring network interfaces...                   [ OK ]
 * Deactivating swap...                                  [fail]
umount: /run/lock: not mounted
mount: cannot mount block device
/dev/disk/by-uuid/9b50a43d-98c3-45ad-a540-7fcbc629a418 read-only
 * Will now halt
#

Any idea about how to investigate/solve this problem??

Thanks in advance,
David Fernandez


-- lxc.xml file --

<domain type='lxc'>
  <name>lxc</name>
  <memory>524288</memory>
  <os>
    <type>exe</type>
    <init>/sbin/init</init>
  </os>
  <vcpu>1</vcpu>
  <clock offset='utc'/>
  <on_poweroff>destroy</on_poweroff>
  <on_reboot>restart</on_reboot>
  <on_crash>destroy</on_crash>
  <devices>
    <emulator>/usr/lib/libvirt/libvirt_lxc</emulator>
    <filesystem type='mount'>
      <source dir='/var/lib/lxc/lxc/rootfs'/>
      <target dir='/'/>
    </filesystem>
    <interface type='network'>
      <source network='default'/>
    </interface>
    <console type='pty'/>
  </devices>
</domain>

On Thu, Jul 05, 2012 at 01:34:00AM +0200, david at dit.upm.es
wrote:
> 
> Hi,
> 
> I've been making some tests with libvirt and LXC and found some
problems
> when halting/restarting a LXC container from within.
> 
> Basically, on a Ubuntu 12.04 system with libvirt installed as package
> (0.9.8), I've created a basic container image with:
> 
>  lxc-create -t ubuntu -n lxc
> 
> And started it using the libvirt XML listed below and the following
command:
> 
>  virsh -c lxc:// create lxc.xml
> 
> I can access the container console normally but when I issue a halt or
> reboot command from inside, the container initiates the halt/reboot
> execution but it does not finishes it properly.
> 
> For example:
> 
> $ sudo halt -p
> [sudo] password for ubuntu:
> 
> Broadcast message from ubuntu at my-container
> 	(/dev/pts/0) at 23:19 ...
> 
> The system is going down for power off NOW!
> ubuntu at my-container:~$ acpid: exiting
>  * Asking all remaining processes to terminate...    [ OK ]
>  * All processes ended within 1 seconds....          [ OK ]
>  * Deconfiguring network interfaces...               [ OK ]
>  * Deactivating swap...                              [fail]
>  * Unmounting weak filesystems...                    [ OK ]
> umount: /run/lock: not mounted
> mount: / is busy
>  * Will now halt
> 
> The container seems to finish the shutdown process but libvirt does not
> seem to be signaled about it (virsh shows the container is still
> executing). Something similar happens with reboot.
> 
> However, if a start that container with:
> 
>  lxc-start -n lxc
> 
> and do the same test, it works perfectly:
> 
> # sudo halt -p
> [sudo] password for ubuntu:
> 
> Broadcast message from ubuntu at my-container
> 	(/dev/lxc/console) at 23:17 ...
> 
> The system is going down for power off NOW!
> ubuntu at my-container:~$ acpid: exiting
>  * Asking all remaining processes to terminate...        [ OK ]
>  * All processes ended within 1 seconds....              [ OK ]
>  * Deconfiguring network interfaces...                   [ OK ]
>  * Deactivating swap...                                  [fail]
> umount: /run/lock: not mounted
> mount: cannot mount block device
> /dev/disk/by-uuid/9b50a43d-98c3-45ad-a540-7fcbc629a418 read-only
>  * Will now halt
> #
> 
> Any idea about how to investigate/solve this problem??
I don't know why it doesn't halt - with recent kernels if you issue
the reboot() syscall from within the guest, then init process should
exit which libvirt will treat as halted. There are patches pending
to make reboots work too but that's for a future release


Daniel
-- 
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|