[ovirt-users] Shutdown Problems on oVirt Node

[Nir Soffer]

On Wed, Mar 2, 2016 at 6:00 PM, Roger Meier <roger.meier at 4synergy.com> wrote:
> Hi their
>
> I have currently a strange problem on a new oVirt 3.6 installation. At
> the moment a clean shutdown doesn't work, most of the time he reboots
> the system or hangs in the shutdown progress.
>
> I discovered this, when i tested our multiple UPS solution and send some
> test signals over ipmi to our server with ex. ipmipower -h 192.168.2.218
> -u root -p password --soft . We also discovered that shutdown -h now ,
> poweroff or init 0 had the same effect.
>
> On a clean CentOS installation, which is not included in our oVirt setup
> this works as expected, but on our ovirt-node this doesn't work.
>
> In the shutdown progress i see the following which tooks very long:
>> A stop job is running for Shared Storage Lease Manager (23s / 1min 47s)

This is sanlock - maybe it would not stop because it has active
lockspaces, delaying shutdown?

Did you put the host to maintenance before shutting it down?

In maintenance mode, vdsm will release all lockspaces, so sanlock
should not delay shutdown in any way.

>
> At the end i had then the following on my console screen:
>
> [ OK ] Reached target Shutdown
>
> Nothing more happens. no poweroff or something. I can wait more than
> three minutes and nothing happens.
>
> I also tried a clean re-install from the oVirt Administration WebUI but
> this doesn't have any effect on this issue.
>
> When i type "service sanlock stop" or "service vdsmd stop" in the server
> console and then do a poweroff , all works as expected. The shutdown is
> the also realy fast as expected.

During poweroff sanlock service is stopped like any other service, so
if it worked from the shell, it should work during shutdown.

Stopping vdsm is not needed and should not effect the shutdown.

> At the moment we think that the problem is on ovirt, vdsmd or on the
> sanlock settings for ovirt, because all settings on our site are on
> default settings.
>
> Currently Setup are two Intel Server (With RRM) with CentOS-7 (1511) and
> oVirt 3.6.3 and one Intel Server with OpenIndiana which provides Storage
> via NFS.
>
> Had someone a solution for this? is this perhaps a bug and shoul'd  be
> reported?

You can file a bug and attach the logs mentioned by Allon, it will help
to track this issue.

Since this is a problem with ovirt-node, I would open a bug for it. It may also
be an issue with sanlock init scripts.

Nir