When deploying a BlockStorage role with an HA Controller the following error will be observed:
Error: pcs -f /var/lib/pacemaker/cib/puppet-cib-backup20180712-8-1yx6b6z create failed: Error: unable to set attribute cinder-volume-role\",
Error: /Stage[main]/Tripleo::Profile::Pacemaker::Cinder::Volume_bundle/Pacemaker::Property[cinder-volume-role-nodefoo]/Pcmk_property[property-nodefoo-cinder-volume-role]/ensure: change from absent to present failed: pcs -f /var/lib/pacemaker/cib/puppet-cib-backup20180712-8-1yx6b6z create failed: Error: unable to set attribute cinder-volume-role",
This problem exists since pike because:
1)
We have a controller role with CinderVolume which will set the service name to 'cinder_volume', which means the controller nodes will contribute populating the 'cinder_volume_short_node_names' hiera list.
2) Then we have the BlockStorage role which defines the non-ha variant of cinder volume called 'OS::TripleO::Services::BlockStorageCinderVolume' which points to puppet/services/cinder-volume.yaml and, since it also has a service_name 'cinder_volume' all the nodes of BlockStorage role will also populate the 'cinder_volume_short_node_names' hiera key
3) The pacemaker implementation of HA services needs to set a cluster per-node property for each node where cinder-volume is allowed to run and uses the 'cinder_volume_short_node_names' for that. Since BlockStorage is in that list but has no pacemaker service attached to it things break down.
The commit that introduced this is the following one:
Cinder volume service is not managed by Pacemaker on BlockStorage
We do not want cinder-volume to be managed by Pacemaker on
BlockStorage nodes, where Pacemaker is not running at all.
This commit + the fact that since pike we started to set all properties from one node by iterating the cinder_volume_short_node_names hiera key breaks this scenario
https:/