Reduce network interruption of quorum migration

We can reduce the potential network connectivity interruption caused by
quorum migration by stopping Keystone and Neutron last and starting them
first, at the expense of longer API downtime (because each kayobe
invocation first generates configuration).

Author: priteau

tools/rabbitmq-quorum-migration.sh

@@ -2,7 +2,8 @@
 
 set -ex
 
-RABBITMQ_SERVICES_TO_RESTART=barbican,blazar,cinder,cloudkitty,designate,heat,ironic,keystone,magnum,manila,neutron,nova,octavia
+RABBITMQ_CRITICAL_SERVICES_TO_RESTART=keystone,neutron
+RABBITMQ_SERVICES_TO_RESTART=barbican,blazar,cinder,cloudkitty,designate,heat,ironic,magnum,manila,nova,octavia
 RABBITMQ_CONTAINER_NAME=rabbitmq
 
 if [[ ! $KAYOBE_CONFIG_PATH ]]; then
@@ -30,6 +31,7 @@ fi
 # Generate new config, stop services using rabbit, and reset rabbit state
 kayobe overcloud service configuration generate --node-config-dir /etc/kolla --kolla-skip-tags rabbitmq-ha-precheck
 kayobe kolla ansible run "stop --yes-i-really-really-mean-it" -kt $RABBITMQ_SERVICES_TO_RESTART
+kayobe kolla ansible run "stop --yes-i-really-really-mean-it" -kt $RABBITMQ_CRITICAL_SERVICES_TO_RESTART
 kayobe kolla ansible run rabbitmq-reset-state
 
 if [[ ! "$1" = "--skip-checks" ]]; then
@@ -47,6 +49,7 @@ if [[ ! "$1" = "--skip-checks" ]]; then
 fi
 
 # Redeploy with quorum queues enabled
+kayobe kolla ansible run deploy-containers -kt $RABBITMQ_CRITICAL_SERVICES_TO_RESTART
 kayobe kolla ansible run deploy-containers -kt $RABBITMQ_SERVICES_TO_RESTART
 
 if [[ ! "$1" = "--skip-checks" ]]; then