Compute node got kernel panic after instance creation

[Artem Panchenko]

Fuel version info (6.1 #98): http://paste.openstack.org/show/167554/

This issue was produced during system tests on CI:

http://jenkins-product.srt.mirantis.net:8080/view/6.1_swarm/job/6.1.system_test.centos.cluster_actions/25/testReport/junit/%28root%29/deploy_stop_reset_on_ha/deploy_stop_reset_on_ha/

OSTF test 'Launch instance, create snapshot, launch instance from snapshot (failure)' failed by timeout, because compute node on which the instance was scheduled to boot got kernel panic (see screenshot) and became inaccessible.

Steps to reproduce:

1. Create new env: CentOS, HA, NovaFlatDHCP, Cinder for volumes
2. Add 3 controllers and 2 computes
3. Deploy changes
4. Run OSTF

Expected result:

- all health checks passed

Actual:

- 'Launch instance, create snapshot, launch instance from snapshot (failure)' test failed by timeout

Here is the part of OSTF logs:

http://paste.openstack.org/show/167564/
http://paste.openstack.org/show/167569/

As you can see from timestamps compute node gone offline when test was running (2015-02-05 05:18). After revert from snapshot crashed node (node-2) was inaccessible via network, VNC console screeshot is attached. When I reset that node it booted correctly and I was able to create instance on it without problems, so I believe that the bug is floating.

Diagnostic snapshot and logs from node-2 are attached.

[Aleksander Mogylchenko (amogylchenko)]

Low priority, because:
- panic occurred in nested virtualization;
- panic occurred only once (according to Artem);

I was able to find only one somewhat similar bug:
https://bugzilla.redhat.com/show_bug.cgi?id=1076294

[Michael Semenov (msemenov)]

Postponing to 7.0 due to low priority.

[Pavel Boldin (pboldin)]

Impossible to fix without debug information.

Please, next time this happens make a dump of the VM memory using QEMU `dump'.