Apply Cinder rolling upgrade

Apply Cinder database migration procedure:

[1] https://docs.openstack.org/cinder/pike/upgrade.html

Depends-On: I9a847b0a916a85d8f2622e93734330da5122f0d8
Change-Id: I934c38cb5366f0b3a65115b4cdd19878efec7ebc

ansible/roles/cinder/defaults/main.yml

@@ -81,6 +81,9 @@ cinder_database_name: "cinder"
 cinder_database_user: "cinder"
 cinder_database_address: "{{ kolla_internal_fqdn }}:{{ database_port }}"
 
+# Max number of object to consider
+# when run online data migration
+# cinder_max_number_osm: ""
 
 ####################
 # Docker

ansible/roles/cinder/tasks/upgrade.yml

@@ -1,7 +1,37 @@
 ---
+# NOTE (duonghq): I break pull and config into 2 phases to prevent new version
+# of service start if node or process in node crash accidentally.
+- include: pull.yml
+  when: inventory_hostname == groups["cinder-api"][0]
+
 - include: config.yml
+  when: inventory_hostname == groups["cinder-api"][0]
 
 - include: bootstrap_service.yml
 
+- include: pull.yml
+
+- include: config.yml
+
 - name: Flush handlers
   meta: flush_handlers
+
+- name: Running Cinder online schema migration
+  vars:
+    cinder_api: "{{ cinder_services['cinder-api'] }}"
+  kolla_docker:
+    action: "start_container"
+    common_options: "{{ docker_common_options }}"
+    detach: False
+    environment:
+      KOLLA_OSM:
+      KOLLA_CONFIG_STRATEGY: "{{ config_strategy }}"
+      MAX_NUMBER: "{{ cinder_max_number_osm | default(10) }}"
+    image: "{{ cinder_api.image }}"
+    labels:
+      BOOTSTRAP:
+    name: "bootstrap_cinder"
+    restart_policy: "never"
+    volumes: "{{ cinder_api.volumes }}"
+  run_once: True
+  delegate_to: "{{ groups[cinder_api.group][0] }}"

releasenotes/notes/implement-cinder-minimal-downtime-upgrade-3c883b015d41bf18.yaml

@@ -0,0 +1,4 @@
+---
+features:
+  - |
+    Implement Cinder minimal downtime upgrade procedure.