TripleO unconditionally calls the openshift-ansible's upgrade playbook on upgrade. It may work fine in the most simplistic case (although likely untested), however this is going to break as soon as more complex deployment are in play, such as with CNS.
In the gluster case case, we have upgrade the control plane and gluster nodes in separate steps and run the playbook on each node, one at a time, after running some manual commands [3]. Perhaps it's something we can automate in TripleO, but more realistically fall under openshift-ansible jurisdiction and as of today, they don't offer an automated solution for gluster nodes.
We should remove the upgrade code from TripleO and give a way to the operator follow the steps outlined in the openshift-ansible's documentation. In other words, we need to provide a way to easily call the openshift-ansible playbooks using the TripleO-generated inventory.
[1] https://github.com/openstack/tripleo-heat-templates/blob/fdba8a8/extraconfig/services/openshift-master.yaml#L63
[2] https://github.com/openstack/tripleo-heat-templates/blob/fdba8a8/extraconfig/services/openshift-master.yaml#L362-L364
[3] https://docs.openshift.com/container-platform/3.11/upgrading/automated_upgrades.html#special-considerations-for-glusterfs
Related fix proposed to branch: master
Review: https:/