[MOS icehouse] Cinder-volume installs on non-cinder nodes

[Tatyanka]

{"build_id": "2014-04-17_15-38-45", "mirantis": "yes", "build_number": "442", "nailgun_sha": "3c49ef0d840d05f258ae88d39656338b0a45c84a", "production": "dev", "ostf_sha": "1ac04769cc8034438ccf1176860b25d0f5601c4a", "fuelmain_sha": "9890d1afdea85343de40fdfb1d91ddab9340082e", "astute_sha": "0cb0b26773bd269000a14748e8ba49419462ecaf", "release": "5.0", "fuellib_sha": "6b948385aad61486a52f3b24609f1d46aa52853e"}

Deployment:
Simple + Ubuntu. Neutron GRE. 1 controller + 2 computes + 1 cinder(for images and volumes).
Networks configs: All net is placed on separate eth

Expected:
Deployment pass, sanity and smoke tests from health check pass too

Actual:
Deployment pass. But volumes are not created.
Command: sudo cinder-rootwrap /etc/cinder/rootwrap.conf env LC_ALL=C vgs --noheadings -o name cinder
Exit code: 5
Stdout: ''
Stderr: ' Volume group "cinder" not found\n'

(See full logs in attachment - node-9 -> controller node-12 cinder)

[Vladimir Kuklin (vkuklin)]

this is due to incorrect logical expression in the manifests: it sets manage_volumes var to true not only for cinder nodes but also for controller nodes. In this case this leads to installation of cinder-volume on the controller node which does not have corresponding volume group.