[Acceptance][8.0] cli_deploy_ceph_neutron: failure when controller deployment

[Vladimir Jigulin]

Steps to reproduce:
1. Create cluster with tun or vlan network
2. Add 3 nodes with controller role
3. Add 2 nodes with compute role
4. Add 2 nodes with ceph role
5. Deploy the cluster
Or you can use cli_deploy_ceph_neutron_{tun or vlan} scenario from fuel-qa

Expected results:
deployment is successful

Actual result:
ERROR [393] Error running RPC method granular_deploy: Deployment failed on nodes 3, trace:
http://paste.openstack.org/show/580974/
puppet logs on node:
/Stage[main]/Galera/Service[mysql] (err): Could not evaluate: Primitive 'p_mysql' was not found in CIB!
http://paste.openstack.org/show/580975/

Reproducibility:
4 times in row on CI
first failure was in 09/10/16 09:43

Snapshot:
https://drive.google.com/file/d/0Bw7ZahkM7_sJclJGM3JPd2FnSnM/view?usp=sharing

[Alexey Stupnikov (astupnikov)]

I have reproduce this test on my virtual environment. It turns out that fuel-qa code [1] configures ceph only as volume and image and volume storage (not ephemeral and object storage). As a result, we catch bug #1604879. It looks like the logs from description are from some sporadic failure and not describing the real problem. Marking this bug as duplicate of bug #1604879.

https://github.com/openstack/fuel-qa/blob/stable-mu/8.0/fuelweb_test/tests/tests_cli/test_cli_deploy_ceph.py#L68