Improve purge playbook

We run a zap twice here since we can't combine with_items and
with_sequence yet. This will land in Ansible 2.0. We destroy all the
partitions and re-create new labels as well. We do this twice just
because sometimes one is not sufficient. We also call partprobe to make
sure that the kernel sees the changes.

Signed-off-by: leseb <seb@redhat.com>

purge-cluster.yml

@@ -12,21 +12,27 @@
 
   tasks:
 
-  - name: Purge Ceph
+  - name: disk zap
+    command: /usr/sbin/sgdisk --zap-all --clear --mbrtogpt -g -- {{ item }}
+    with_items: devices
+    ignore_errors: true
+
+  - name: disk zap
+    command: /usr/sbin/sgdisk --zap-all --clear --mbrtogpt -g -- {{ item }}
+    with_items: devices
+    ignore_errors: true
+
+  - name: call partprobe
+    command: partprobe
+
+  - name: purge ceph
     command: ceph-deploy purge {{ ansible_fqdn }}
     delegate_to: 127.0.0.1
 
-  - name: Remove OSD data
+  - name: remove osd data
     shell: rm -rf /var/lib/ceph/osd/*/*
     ignore_errors: true
 
-  - name: Purge remaining data
+  - name: purge remaining data
     command: ceph-deploy purgedata {{ ansible_fqdn }}
     delegate_to: 127.0.0.1
-
-  - name: Purge partitions
-    shell: parted -s {{ item[0] }} rm {{ item[1] }}
-    with_nested:
-      - devices
-      - partitions
-    ignore_errors: true