[Acceptance][8.0] OSTF failed in test ceph_volumes_ephemeral

[Vladimir Jigulin]

Reproduced on CI: https://patching-ci.infra.mirantis.net/job/8.0.acceptance.ubuntu.ha_vlan_group_3/4/testReport/(root)/ceph_volumes_ephemeral/ceph_volumes_ephemeral/

Steps to reproduce:
1. Create new environment
2. Choose Neutron, VLAN
3. Choose Ceph for volumes and Ceph for ephemeral
4. Change openstack username, password, tenant
5. Add 3 controller
6. Add 2 compute
7. Add 3 ceph nodes
8. Change default management net mask from /24 to /25
9. Verify networks
10. Start deployment
11. Verify networks
12. Run OSTF

or use ceph_volumes_ephemeraltest from fuel-qa
test fail in line: https://github.com/openstack/fuel-qa/blob/stable/8.0/fuelweb_test/tests/tests_deployments/tests_neutron_vlan/test_ha_vlan_group_3.py#L175

Expected results:
OSTF test should be passed

Actual result:
Failed 1 OSTF tests; Names of failed tests:
 - Instance live migration (failure) VM connectivity doesn`t function properly.

Reproducibility: rarely

Snapshot: https://drive.google.com/file/d/0Bw7ZahkM7_sJeV9LbXlTWjhCRms/view?usp=sharing

[Vladimir Jigulin (vjigulin)]

Probably we have same reason in periodic failure in some "ceph ephemeral" tests:
ceph_for_images_ephemeral_rados
ceph_for_volumes_images_ephemeral_rados
tun_no_volumes_ceph_for_images_and_ephemeral

Snapshots:
https://drive.google.com/file/d/0Bw7ZahkM7_sJOWNPbzl4cGVZMG8/view?usp=sharing
https://drive.google.com/file/d/0Bw7ZahkM7_sJUTVHVUJJOTVjTnc/view?usp=sharing
https://drive.google.com/file/d/0Bw7ZahkM7_sJeV9LbXlTWjhCRms/view?usp=sharing

[Rodion Tikunov (rtikunov)]

Reproduced in the test lab. The root cause of this bug that sometimes (about once in 5-6 times) VM does not boot after live migration. Core dump appears in console logs. But after hard rebooting this VM is booted normally and floating IP is pinged.
Tested with TestVM image (kernel 3.2.0-80-virtual) and lastest cirros images [0] (kernel 4.4.0-28-generic).

Console log from out-of-box TestVM [1]
Console log from lastest Cirros image [2]

[0] http://download.cirros-cloud.net/daily/20160722/cirros-d160722-i386-disk.img
[1] http://paste.openstack.org/show/563916/
[2] http://paste.openstack.org/show/563906/

[Denis Meltsaykin (dmeltsaykin)]

Removing from 8.0-mu-3 as the fix is not available.

[Rodion Tikunov (rtikunov)]

Run the script on controller node.