[train-only] Ensure we stop ovn-controller with cleanup

Turns out that gracefull shutdown of container running ovn_controller
does not disable routing to it. This means that after Leapp when the
node is clean we still try to send traffic to it and this causes
outage. By calling ovs-appctl -t ovn-controller exit we ensure that
ovn cleans up BFD interfaces from the node. For this to work we need
to disable docker autorestart of the container.

Change-Id: I9ec73e3d94884c339ea5e21817c24cf5e758f242

deployment/ovn/ovn-controller-container-puppet.yaml

@@ -318,6 +318,28 @@ outputs:
             persistent: yes
             state: yes
       upgrade_tasks:
+        # system_upgrade
+        - name: ovn_controller system_upgrade_prepare step 1
+          tags:
+            - never
+            - system_upgrade
+            - system_upgrade_prepare
+          when:
+            - step|int == 1
+          block:
+            - name: Check ovn_controller is running in docker
+              shell: |
+                docker ps | grep ovn_controller
+              register: ovn_controller_running
+              failed_when: false
+            - name: Disable autorestart on ovn_controller container
+              command: docker update --restart=no ovn_controller
+              when: ovn_controller_running.rc == 0
+            - name: Tell ovn_controller to clean up and stop
+              shell: |
+                docker exec -u root ovn_controller bash -c "if [ -f /usr/bin/ovn-appctl ] ; then ovn-appctl -t ovn-controller exit ; else ovs-appctl -t ovn-controller exit ; fi"
+              when: ovn_controller_running.rc == 0
+        # nova_hybrid_state
         - name: Gather missing facts
           setup:
             gather_subset: "distribution"