Merge pull request #166 from stackhpc/stable/victoria

sync stackhpc/victoria with upstream

Author: bbezak

ansible/gather-facts.yml

@@ -17,6 +17,7 @@
     - name: Group hosts to determine when using --limit
       group_by:
         key: "all_using_limit_{{ (ansible_play_batch | length) != (groups['all'] | length) }}"
+      changed_when: false
   tags: always
 
 # NOTE(pbourke): This case covers deploying subsets of hosts using --limit. The

ansible/group_vars/all.yml

@@ -538,7 +538,7 @@ admin_protocol: "{{ 'https' if kolla_enable_tls_internal | bool else 'http' }}"
 openstack_release: "victoria"
 # Docker image tag used by default.
 openstack_tag: "{{ openstack_release ~ openstack_tag_suffix }}"
-openstack_tag_suffix: ""
+openstack_tag_suffix: "{{ '-centos8s' if 'Stream' in (kolla_centos_release['content'] | default() | b64decode ) else '' }}"
 openstack_logging_debug: "False"
 
 openstack_region_name: "RegionOne"
@@ -1074,7 +1074,7 @@ ceph_cinder_user: "cinder"
 ceph_glance_user: "glance"
 ceph_gnocchi_user: "gnocchi"
 ceph_manila_user: "manila"
-ceph_nova_user: "nova"
+ceph_nova_user: "{{ ceph_cinder_user }}"
 
 # External Ceph keyrings
 ceph_cinder_keyring: "ceph.client.cinder.keyring"
@@ -1122,6 +1122,7 @@ enable_prometheus_rabbitmq_exporter: "{{ enable_prometheus | bool and enable_rab
 
 prometheus_alertmanager_user: "admin"
 prometheus_openstack_exporter_interval: "60s"
+prometheus_openstack_exporter_timeout: "10s"
 prometheus_elasticsearch_exporter_interval: "60s"
 prometheus_cmdline_extras:
 prometheus_ceph_mgr_exporter_endpoints: []

ansible/library/kolla_docker.py

@@ -247,6 +247,9 @@
 '''
 
 
+COMPARE_CONFIG_CMD = ['/usr/local/bin/kolla_set_configs', '--check']
+
+
 def get_docker_client():
     return docker.APIClient
 
@@ -328,7 +331,9 @@ def get_container_info(self):
 
     def compare_container(self):
         container = self.check_container()
-        if not container or self.check_container_differs():
+        if (not container or
+                self.check_container_differs() or
+                self.compare_config()):
             self.changed = True
         return self.changed
 
@@ -566,6 +571,43 @@ def compare_healthcheck(self, container_info):
             if current_healthcheck:
                 return True
 
+    def compare_config(self):
+        try:
+            job = self.dc.exec_create(
+                self.params['name'],
+                COMPARE_CONFIG_CMD,
+                user='root',
+            )
+            output = self.dc.exec_start(job)
+            exec_inspect = self.dc.exec_inspect(job)
+        except docker.errors.APIError as e:
+            # NOTE(yoctozepto): If we have a client error, then the container
+            # cannot be used for config check (e.g., is restarting, or stopped
+            # in the mean time) - assume config is stale = return True.
+            # Else, propagate the server error back.
+            if e.is_client_error():
+                return True
+            else:
+                raise
+        # Exit codes:
+        # 0: not changed
+        # 1: changed
+        # 137: abrupt exit -> changed
+        # else: error
+        if exec_inspect['ExitCode'] == 0:
+            return False
+        elif exec_inspect['ExitCode'] == 1:
+            return True
+        elif exec_inspect['ExitCode'] == 137:
+            # NOTE(yoctozepto): This is Docker's command exit due to container
+            # exit. It means the container is unstable so we are better off
+            # marking it as requiring a restart due to config update.
+            return True
+        else:
+            raise Exception('Failed to compare container configuration: '
+                            'ExitCode: %s Message: %s' %
+                            (exec_inspect['ExitCode'], output))
+
     def parse_image(self):
         full_image = self.params.get('image')
 

ansible/roles/common/tasks/config.yml

@@ -277,6 +277,7 @@
   become: true
   when:
     - common_services['kolla-toolbox'] | service_enabled_and_mapped_to_host
+    - enable_rabbitmq | bool
     - api_address_family == "ipv6"
 
 - name: Copy rabbitmq erl_intr to kolla toolbox
@@ -288,4 +289,5 @@
   become: true
   when:
     - common_services['kolla-toolbox'] | service_enabled_and_mapped_to_host
+    - enable_rabbitmq | bool
     - api_address_family == "ipv6"

ansible/roles/common/tasks/inspect.yml

@@ -6,7 +6,7 @@
             "dest": "/var/lib/rabbitmq/.erlang.cookie",
             "owner": "rabbitmq",
             "perm": "0600"
-        }{% endif %}{% if api_address_family == 'ipv6' %},
+        }{% if api_address_family == 'ipv6' %},
         {
             "source": "{{ container_config_directory }}/rabbitmq-env.conf",
             "dest": "/etc/rabbitmq/rabbitmq-env.conf",
@@ -18,7 +18,7 @@
             "dest": "/etc/rabbitmq/erl_inetrc",
             "owner": "rabbitmq",
             "perm": "0600"
-        }{% endif %}
+        }{% endif %}{% endif %}
     ],
     "permissions": [
         {

ansible/roles/common/templates/kolla-toolbox.json.j2

@@ -24,6 +24,7 @@
     re_safe_address: "{{ kolla_internal_vip_address | regex_escape }}"
   group_by:
     key: kolla_ha_is_master_{{ ip_addr_output.stdout is regex('\b' + re_safe_address + '\b') }}
+  changed_when: false
   when:
     - kolla_action != "config"
   listen:

ansible/roles/haproxy/handlers/main.yml

@@ -16,13 +16,15 @@
 - name: Group hosts by whether they are running keepalived
   group_by:
     key: "keepalived_running_{{ container_facts['keepalived'] is defined }}"
+  changed_when: false
   when:
     - enable_keepalived | bool
     - inventory_hostname in groups['haproxy']
 
 - name: Group hosts by whether they are running HAProxy
   group_by:
     key: "haproxy_running_{{ container_facts['haproxy'] is defined }}"
+  changed_when: false
   when:
     - enable_haproxy | bool
     - inventory_hostname in groups['haproxy']

ansible/roles/haproxy/tasks/precheck.yml

@@ -14,6 +14,7 @@
 - name: Group nodes where keystone_fernet is running
   group_by:
     key: keystone_fernet_{{ container_facts['keystone_fernet'].State | default('bootstrap') }}
+  changed_when: false
 
 # NOTE(mgoddard): If we bootstrap Fernet keys on an existing cluster, this
 # would overwrite existing keys, and invalidate tokens created from them.

ansible/roles/keystone/tasks/bootstrap_service.yml

@@ -10,6 +10,7 @@
 - name: Divide hosts by their MariaDB volume availability
   group_by:
     key: mariadb_had_volume_{{ mariadb_volume is not changed }}
+  changed_when: false
 
 - name: Establish whether the cluster has already existed
   set_fact:
@@ -29,6 +30,7 @@
     - name: Divide hosts by their MariaDB service port liveness
       group_by:
         key: mariadb_port_alive_{{ check_mariadb_port_liveness is success }}
+      changed_when: false
 
     - name: Fail on existing but stopped cluster
       fail:
@@ -64,6 +66,7 @@
     - name: Divide hosts by their MariaDB service WSREP sync status
       group_by:
         key: mariadb_sync_status_{{ mariadb_sync_status | default('NA') }}
+      changed_when: false
 
     - name: Fail when MariaDB services are not synced across the whole cluster
       fail: