[mos] Ceilometer test to check the alarm can change status failed according to metrics are not created

Bug is reproduced on ISO#235

Steps:
1. Create next env - Centos, HA, KVM, Nova-Network with Vlan, Ceph for volumes and Ceph for images, Ceilometer service
2. Add 3 controllers, 1 compute, 3 Ceph nodes
3. Deploy cluster - deployment is successful
4. Open Health Check tab and run 'Ceilometer test to check the alarm can change status'

Actual result - test failed on step 3. Create metrics for instance. - with timeout error

5. After that allocate more memory for compute and all controller nodes and restart these nodes

Actual result - test failed on step 2. Instance become active. with timeout error

Diagnostic snapshot is attached

Get next error in ceilometer-all.log
libvirtError: Requested operation is not valid: domain is not running
<0>Mar 6 09:03:48 node-4 �<44>ceilometer-ceilometer.compute.pollsters.disk WARNING: Ignoring instance instance-00000021: Requested operation is not valid: domain is not running
<43>Mar 6 09:03:48 node-4 ceilometer-ceilometer.compute.pollsters.disk ERROR: Requested operation is not valid: domain is not running
Traceback (most recent call last):
  File "/usr/lib/python2.6/site-packages/ceilometer/compute/pollsters/disk.py", line 87, in get_samples
    instance_name,
  File "/usr/lib/python2.6/site-packages/ceilometer/compute/pollsters/disk.py", line 59, in _populate_cache
    for disk, info in inspector.inspect_disks(instance_name):
  File "/usr/lib/python2.6/site-packages/ceilometer/compute/virt/libvirt/inspector.py", line 150, in inspect_disks
    block_stats = domain.blockStats(device)
  File "/usr/lib64/python2.6/site-packages/libvirt.py", line 1820, in blockStats
    if ret is None: raise libvirtError ('virDomainBlockStats() failed', dom=self)