NSX|v: Support more than 2 hostgroups

If edge_ha=True and ha_placement_random=True
The user can configure more than 2 edge_host_groups globally or
per availability zone.
In this case 2 of those will be randomly selected for each deployed edge.

Change-Id: Iaa673e5acf78ecdf8cbc942499cce70a2fe0546c

vmware_nsx/common/config.py

@@ -484,8 +484,8 @@ nsxv_opts = [
                 default=[],
                 help=_('(Optional) If edge HA is used then this will ensure '
                        'that active/backup edges are placed in the listed '
-                       'host groups. 2 predefined host groups need to be '
+                       'host groups. At least 2 predefined host groups need '
-                       'configured.')),
+                       'to be configured.')),
     cfg.StrOpt('external_network',
                help=_('(Required) Network ID for physical network '
                       'connectivity')),
@@ -709,8 +709,8 @@ nsxv_az_opts = [
                 default=[],
                 help=_('(Optional) If edge HA is used then this will ensure '
                        'that active/backup edges are placed in the listed '
-                       'host groups. 2 predefined host groups need to be '
+                       'host groups. At least 2 predefined host groups need '
-                       'configured.')),
+                       'to be configured.')),
     cfg.StrOpt('datacenter_moid',
                help=_('(Optional) Identifying the ID of datacenter to deploy '
                       'NSX Edges')),

vmware_nsx/dvs/dvs.py

@@ -161,7 +161,6 @@ class DvsManager(VCManagerBase):
                   'vlan_tag': vlan_tag,
                   'dvs': dvs_moref.value})
 
-    # DEBUG ADIT used only by the DVS plugin
     def _net_id_to_moref(self, dvs_moref, net_id):
         """Gets the moref for the specific neutron network."""
         # NOTE(garyk): return this from a cache if not found then invoke
@@ -564,9 +563,9 @@ class ClusterManager(VCManagerBase):
             reconfig_task = session.invoke_api(
                 session.vim, "ReconfigureComputeResource_Task",
                 cluster, spec=config_spec, modify=True)
+            session.wait_for_task(reconfig_task)
         except Exception as excep:
             LOG.exception('Failed to reconfigure cluster %s', excep)
-        session.wait_for_task(reconfig_task)
 
     def _create_vm_group_spec(self, client_factory, name, vm_refs,
                               group=None):
@@ -605,7 +604,7 @@ class ClusterManager(VCManagerBase):
         rules_spec.info = rules_info
         return rules_spec
 
-    def get_configured_vms(self, resource_id):
+    def get_configured_vms(self, resource_id, n_host_groups=2):
         session = self._session
         resource = vim_util.get_moref(resource_id, 'ResourcePool')
         # TODO(garyk): cache the cluster details
@@ -616,7 +615,7 @@ class ClusterManager(VCManagerBase):
             vim_util, "get_object_property", self._session.vim, cluster,
             "configurationEx")
         configured_vms = []
-        for index in range(2):
+        for index in range(n_host_groups):
             vm_group = None
             entry_id = index + 1
             groups = []
@@ -632,8 +631,9 @@ class ClusterManager(VCManagerBase):
         return configured_vms
 
     def update_cluster_edge_failover(self, resource_id, vm_moids,
-                                     edge_id, host_group_names):
+                                     host_group_names):
         """Updates cluster for vm placement using DRS"""
+        # DEBUG ADIT edge-id is never used
         session = self._session
         resource = vim_util.get_moref(resource_id, 'ResourcePool')
         # TODO(garyk): cache the cluster details
@@ -643,12 +643,20 @@ class ClusterManager(VCManagerBase):
         cluster_config = session.invoke_api(
             vim_util, "get_object_property", self._session.vim, cluster,
             "configurationEx")
-        vms = [vim_util.get_moref(vm_moid, 'VirtualMachine') for
+        vms = [vim_util.get_moref(vm_moid, 'VirtualMachine')
-               vm_moid in vm_moids]
+               if vm_moid else None
+               for vm_moid in vm_moids]
         client_factory = session.vim.client.factory
         config_spec = client_factory.create('ns0:ClusterConfigSpecEx')
         num_host_groups = len(host_group_names)
+
+        rules = []
+        if hasattr(cluster_config, 'rule'):
+            rules = cluster_config.rule
+
         for index, vm in enumerate(vms, start=1):
+            if not vm:
+                continue
             vmGroup = None
             groups = []
             if hasattr(cluster_config, 'group'):
@@ -664,9 +672,6 @@ class ClusterManager(VCManagerBase):
                             [vm], vmGroup)
             config_spec.groupSpec.append(groupSpec)
             config_rule = None
-            rules = []
-            if hasattr(cluster_config, 'rule'):
-                rules = cluster_config.rule
             # Create the config rule if it does not exist
             for rule in rules:
                 if 'neutron-rule-%s' % index == rule.name:
@@ -691,11 +696,11 @@ class ClusterManager(VCManagerBase):
         cluster_config = session.invoke_api(
             vim_util, "get_object_property", self._session.vim, cluster,
             "configurationEx")
-        for host_group_name in host_group_names:
-            found = False
         groups = []
         if hasattr(cluster_config, 'group'):
             groups = cluster_config.group
+        for host_group_name in host_group_names:
+            found = False
             for group in groups:
                 if host_group_name == group.name:
                     found = True
@@ -703,15 +708,16 @@ class ClusterManager(VCManagerBase):
             if not found:
                 LOG.error("%s does not exist", host_group_name)
                 raise exceptions.NotFound()
+
         update_cluster = False
         num_host_groups = len(host_group_names)
+        rules = []
+        if hasattr(cluster_config, 'rule'):
+            rules = cluster_config.rule
         # Ensure that the VM groups are created
-        for index in range(2):
+        for index in range(num_host_groups):
             entry_id = index + 1
             vmGroup = None
-            groups = []
-            if hasattr(cluster_config, 'group'):
-                groups = cluster_config.group
             for group in groups:
                 if 'neutron-group-%s' % entry_id == group.name:
                     vmGroup = group
@@ -725,9 +731,6 @@ class ClusterManager(VCManagerBase):
                 update_cluster = True
 
             config_rule = None
-            rules = []
-            if hasattr(cluster_config, 'rule'):
-                rules = cluster_config.rule
             # Create the config rule if it does not exist
             for rule in rules:
                 if 'neutron-rule-%s' % entry_id == rule.name:
@@ -744,8 +747,7 @@ class ClusterManager(VCManagerBase):
             try:
                 self._reconfigure_cluster(session, cluster, config_spec)
             except Exception as e:
-                LOG.error('Unable to update cluster for host groups %s',
+                LOG.error('Unable to update cluster for host groups %s', e)
-                          e)
 
     def _delete_vm_group_spec(self, client_factory, name):
         group_spec = client_factory.create('ns0:ClusterGroupSpec')
@@ -768,7 +770,7 @@ class ClusterManager(VCManagerBase):
         rules_spec.info = rules_info
         return rules_spec
 
-    def cluster_host_group_cleanup(self, resource_id):
+    def cluster_host_group_cleanup(self, resource_id, n_host_groups=2):
         session = self._session
         resource = vim_util.get_moref(resource_id, 'ResourcePool')
         # TODO(garyk): cache the cluster details
@@ -780,21 +782,22 @@ class ClusterManager(VCManagerBase):
         cluster_config = session.invoke_api(
             vim_util, "get_object_property", self._session.vim, cluster,
             "configurationEx")
-        groupSpec = []
-        ruleSpec = []
-        for index in range(2):
-            entry_id = index + 1
         groups = []
         if hasattr(cluster_config, 'group'):
             groups = cluster_config.group
+        rules = []
+        if hasattr(cluster_config, 'rule'):
+            rules = cluster_config.rule
+
+        groupSpec = []
+        ruleSpec = []
+        for index in range(n_host_groups):
+            entry_id = index + 1
             for group in groups:
                 if 'neutron-group-%s' % entry_id == group.name:
                     groupSpec.append(self._delete_vm_group_spec(
                         client_factory, group.name))
-            rules = []
+            # Delete the config rule if it exists
-            if hasattr(cluster_config, 'rule'):
-                rules = cluster_config.rule
-            # Create the config rule if it does not exist
             for rule in rules:
                 if 'neutron-rule-%s' % entry_id == rule.name:
                     ruleSpec.append(self._delete_cluster_rules_spec(

vmware_nsx/plugins/nsx_v/plugin.py

@@ -4193,6 +4193,18 @@ class NsxVPluginV2(addr_pair_db.AllowedAddressPairsMixin,
             azs = self.get_azs_list()
             for az in azs:
                 if az.edge_host_groups and az.edge_ha:
+                    if len(az.edge_host_groups) < 2:
+                        error = _("edge_host_groups must have at least 2 "
+                                  "names")
+                        raise nsx_exc.NsxPluginException(err_msg=error)
+                    if (not az.ha_placement_random and
+                        len(az.edge_host_groups) > 2):
+                        LOG.warning("Availability zone %(az)s has %(count)s "
+                                    "hostgroups. only the first 2 will be "
+                                    "used until ha_placement_random is "
+                                    "enabled",
+                                    {'az': az.name,
+                                     'count': len(az.edge_host_groups)})
                     self._vcm.validate_host_groups(az.resource_pool,
                                                    az.edge_host_groups)
 

vmware_nsx/plugins/nsx_v/vshield/edge_utils.py

@@ -723,7 +723,6 @@ class EdgeManager(object):
                     context, edge_id, lrouter['name'], lrouter['id'], dist,
                     True, availability_zone=availability_zone,
                     deploy_metadata=deploy_metadata)
-
                 try:
                     self.nsxv_manager.rename_edge(edge_id, name)
                 except nsxapi_exc.VcnsApiException as e:
@@ -2603,22 +2602,26 @@ def update_edge_host_groups(vcns, edge_id, dvs, availability_zone,
            for appliance in appliances['appliances']]
     if validate:
         configured_vms = dvs.get_configured_vms(
-            availability_zone.resource_pool)
+            availability_zone.resource_pool,
+            len(availability_zone.edge_host_groups))
         for vm in vms:
             if vm in configured_vms:
                 LOG.info('Edge %s already configured', edge_id)
                 return
+    LOG.info('Create DRS groups for %(vms)s on edge %(edge_id)s',
+             {'vms': vms, 'edge_id': edge_id})
     # Ensure random distribution of the VMs
     if availability_zone.ha_placement_random:
+        if len(vms) < len(availability_zone.edge_host_groups):
+            # add some empty vms to the list, so it will randomize between
+            # all host groups
+            vms.extend([None] * (len(availability_zone.edge_host_groups) -
+                                 len(vms)))
         random.shuffle(vms)
     try:
-        LOG.info('Create DRS groups for '
-                 '%(vms)s on edge %(edge_id)s',
-                 {'vms': vms,
-                  'edge_id': edge_id})
         dvs.update_cluster_edge_failover(
             availability_zone.resource_pool,
-            vms, edge_id, availability_zone.edge_host_groups)
+            vms, availability_zone.edge_host_groups)
     except Exception as e:
         LOG.error('Unable to create DRS groups for '
                   '%(vms)s on edge %(edge_id)s. Error: %(e)s',
@@ -2632,7 +2635,8 @@ def clean_host_groups(dvs, availability_zone):
         LOG.info('Cleaning up host groups for AZ %s',
                  availability_zone.name)
         dvs.cluster_host_group_cleanup(
-            availability_zone.resource_pool)
+            availability_zone.resource_pool,
+            len(availability_zone.edge_host_groups))
     except Exception as e:
         LOG.error('Unable to cleanup. Error: %s', e)