libvirt users - Sep 2013 - Re: Unable to start libvirt_lxc domain a second time after shutdown

Hi,

I'm unable to start a lxc domain again after shutting it down.

Something doesn't seem to "clean up" the socket file (and probably
more), as the error message says

"Unable to start Domain: Failed to connect socket to
'/var/run/libvirt/lxc/test.sock': Connection refused."

I can create a domain with the same config as the one thats shutted
down, but with other uid and name, of course, and I'm able to boot up
this domain.

However, the init process is much shorter (some filesystems are not
mounted), so maybe there is still something left "open" after the lxc
container is shut down?

I saw the problem in libvirt 1.1.2 and current git (1.1.3-rc2 plus 7
commits, on 51e21ba7b6f59d6ad9339b2da2b75cc7704f9ecf)


Configuration is something like this:

<domain type='lxc'>
  <name>test</name>
  <memory unit='KiB'>1048576</memory>
  <currentMemory unit='KiB'>1048576</currentMemory>
  <vcpu placement='static'>1</vcpu>
  <resource>
    <partition>/machine</partition>
  </resource>
  <os>
    <type arch='x86_64'>exe</type>
    <init>/sbin/init</init>
  </os>
  <clock offset='utc'/>
  <on_poweroff>destroy</on_poweroff>
  <on_reboot>restart</on_reboot>
  <on_crash>restart</on_crash>
  <devices>
    <emulator>/usr/lib/libvirt/libvirt_lxc</emulator>
    <filesystem type='mount' accessmode='passthrough'>
      <source dir='/var/lib/libvirt/filesystems/test/root'/>
      <target dir='/'/>
    </filesystem>
    <interface type='network'>
    </interface>
    <console type='pty'>
      <target type='lxc' port='0'/>
    </console>
  </devices>
  <seclabel type='none'/>
</domain>


Florian

On Mon, Sep 30, 2013 at 04:42:18PM +0200, Florian Klink
wrote:
> Hi,
> 
> I'm unable to start a lxc domain again after shutting it down.
> 
> Something doesn't seem to "clean up" the socket file (and
probably
> more), as the error message says
> 
> "Unable to start Domain: Failed to connect socket to
> '/var/run/libvirt/lxc/test.sock': Connection refused."
> 
> I can create a domain with the same config as the one thats shutted
> down, but with other uid and name, of course, and I'm able to boot up
> this domain.
> 
> However, the init process is much shorter (some filesystems are not
> mounted), so maybe there is still something left "open" after the
lxc
> container is shut down?
> 
> I saw the problem in libvirt 1.1.2 and current git (1.1.3-rc2 plus 7
> commits, on 51e21ba7b6f59d6ad9339b2da2b75cc7704f9ecf)
I'm assuming you are running on a distro with systemd ?

If so, can you try out this patch

  https://www.redhat.com/archives/libvir-list/2013-September/msg01671.html

It tries to workaround a flaw in either systemd or kernel (we're not
sure which is broken) that causes cgroups to not be cleaned up

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 :|

Am 30.09.2013 16:44, schrieb Daniel P. Berrange:
> On Mon, Sep 30, 2013 at 04:42:18PM +0200, Florian Klink wrote:
>> Hi,
>>
>> I'm unable to start a lxc domain again after shutting it down.
>>
>> Something doesn't seem to "clean up" the socket file (and
probably
>> more), as the error message says
>>
>> "Unable to start Domain: Failed to connect socket to
>> '/var/run/libvirt/lxc/test.sock': Connection refused."
>>
>> I can create a domain with the same config as the one thats shutted
>> down, but with other uid and name, of course, and I'm able to boot
up
>> this domain.
>>
>> However, the init process is much shorter (some filesystems are not
>> mounted), so maybe there is still something left "open" after
the lxc
>> container is shut down?
>>
>> I saw the problem in libvirt 1.1.2 and current git (1.1.3-rc2 plus 7
>> commits, on 51e21ba7b6f59d6ad9339b2da2b75cc7704f9ecf)
> I'm assuming you are running on a distro with systemd ?
>
> If so, can you try out this patch
>
>   https://www.redhat.com/archives/libvir-list/2013-September/msg01671.html
>
> It tries to workaround a flaw in either systemd or kernel (we're not
> sure which is broken) that causes cgroups to not be cleaned up
>
> Daniel
Bingo! This one worked around the issue, thanks!

Are you considering to add something like this workaround to the
libvirt-1.1.3 release?
I'm pretty sure I'm not the only one affected by this... ;-)

Florian