Make the metalsmith module interactions more resilient

This change adds retries around the metalsmith module interactions.
These retries will ensure greater degrees of success when dealing with
potentially flaky API calls.

Change-Id: Ibc5d34f5322340e91163830426fa400c990c48f9
Signed-off-by: Kevin Carter <kecarter@redhat.com>

tripleo_ansible/playbooks/cli-overcloud-node-provision.yaml

@@ -102,6 +102,9 @@
         state: reserved
         clean_up: true
       register: baremetal_reserved
+      until: baremetal_reserved is success
+      retries: 3
+      delay: 2
 
     - name: Provision instances
       metalsmith_instances:
@@ -112,6 +115,9 @@
         timeout: "{{ node_timeout }}"
         concurrency: "{{ concurrency }}"
       register: baremetal_provisioned
+      until: baremetal_provisioned is success
+      retries: 3
+      delay: 2
 
     - name: Populate environment
       tripleo_baremetal_populate_environment:

tripleo_ansible/playbooks/cli-overcloud-node-unprovision.yaml

@@ -79,3 +79,7 @@
         instances: "{{ baremetal_existing.instances }}"
         state: absent
       when: not prompt|bool
+      register: baremetal_unprovisioned
+      until: baremetal_unprovisioned is success
+      retries: 3
+      delay: 2