X-Git-Url: https://gerrit.opnfv.org/gerrit/gitweb?a=blobdiff_plain;f=puppet%2Fservices%2Fceph-mon.yaml;h=45fca89f588222a75e0014112ca11dd2757becaf;hb=e4c07e2ab055481a0e3986122eca499659aebd33;hp=1ce5833529107d4f0f38a93e34b5521101e7db33;hpb=854b5120cc04418b34e45a523d0ea67ae64a448e;p=apex-tripleo-heat-templates.git

diff --git a/puppet/services/ceph-mon.yaml b/puppet/services/ceph-mon.yaml
index 1ce58335..45fca89f 100644
--- a/puppet/services/ceph-mon.yaml
+++ b/puppet/services/ceph-mon.yaml
@@ -13,6 +13,14 @@ parameters:
   DefaultPasswords:
     default: {}
     type: json
+  RoleName:
+    default: ''
+    description: Role name on which the service is applied
+    type: string
+  RoleParameters:
+    default: {}
+    description: Parameters specific to the role
+    type: json
   EndpointMap:
     default: {}
     description: Mapping of service endpoint -> protocol. Typically set
@@ -59,6 +67,14 @@ parameters:
       }
     default: {}
     type: json
+  CephValidationRetries:
+    type: number
+    default: 5
+    description: Number of retry attempts for Ceph validation
+  CephValidationDelay:
+    type: number
+    default: 10
+    description: Interval (in seconds) in between validation checks
   MonitoringSubscriptionCephMon:
     default: 'overcloud-ceph-mon'
     type: string
@@ -70,6 +86,8 @@ resources:
       ServiceNetMap: {get_param: ServiceNetMap}
       DefaultPasswords: {get_param: DefaultPasswords}
       EndpointMap: {get_param: EndpointMap}
+      RoleName: {get_param: RoleName}
+      RoleParameters: {get_param: RoleParameters}
 
 outputs:
   role_data:
@@ -119,21 +137,32 @@ outputs:
         # rolling upgrade of all osd nodes in step1
         - name: Check status
           tags: step0,validation
-          shell: ceph health | grep -qv HEALTH_ERR
-        # FIXME(shardy) I suspect we can use heat or ansible facts here instead?
-        - name: Get hostname
-          tags: step0
-          shell: hostname -s
-          register: mon_id
-        - name: Stop Ceph Mon
+          shell: ceph health | egrep -sq "HEALTH_OK|HEALTH_WARN"
+        - name: Stop CephMon
           tags: step0
-          service: name=ceph-mon@{{mon_id.stdout}} pattern=ceph-mon state=stopped
-        - name: Update ceph packages
+          service:
+            name: ceph-mon@{{ ansible_hostname }}
+            state: stopped
+        - name: Update Ceph packages
           tags: step0
-          yum: name=ceph-mon state=latest
-        - name: Start ceph-mon service
+          yum:
+            name: ceph-mon
+            state: latest
+        - name: Start CephMon
           tags: step0
-          service: name=ceph-mon@{{mon_id.stdout}} state=started
+          service:
+            name: ceph-mon@{{ ansible_hostname }}
+            state: started
+        # ceph-ansible
+        # https://github.com/ceph/ceph-ansible/blob/master/infrastructure-playbooks/rolling_update.yml#L149-L157
+        - name: Wait for the monitor to join the quorum...
+          tags: step0,ceph_quorum_validation
+          shell: |
+            ceph -s | grep monmap | sed 's/.*quorum//' | egrep -sq {{ ansible_hostname }}
+          register: ceph_quorum_nodecheck
+          until: ceph_quorum_nodecheck.rc == 0
+          retries: {get_param: CephValidationRetries}
+          delay: {get_param: CephValidationDelay}
         - name: ceph osd crush tunables default
           tags: step0
           shell: ceph osd crush tunables default