aggregation fix

watcher/decision_engine/datasources/prometheus_base.py

@@ -40,8 +40,8 @@ class PrometheusBase(base.DataSourceBase):
                       host_inlet_temp=None,
                       host_airflow=None,
                       host_power=None,
-                      instance_cpu_usage='ceilometer_cpu',
-                      instance_ram_usage='ceilometer_memory_usage',
+                      instance_cpu_usage='libvirt_domain_info_cpu_time_seconds_total',
+                      instance_ram_usage='libvirt_domain_info_memory_usage_bytes',
                       instance_ram_allocated='instance.memory',
                       instance_l3_cache_usage=None,
                       instance_root_disk_size='instance.disk',
@@ -156,7 +156,7 @@ class PrometheusBase(base.DataSourceBase):
             host: fqdn for (host, fqdn) in (
                 (fqdn.split('.')[0], fqdn)
                 for fqdn in self.prometheus_fqdn_labels
-                if '.' in fqdn
+                # if '.' in fqdn
             )
         }
         if not host_instance_map:
@@ -311,6 +311,18 @@ class PrometheusBase(base.DataSourceBase):
                    'agg': aggregate, 'meter': meter, 'period': period,
                    'vcpus': vcpus}
             )
+        elif meter == 'libvirt_domain_info_cpu_time_seconds_total':
+            query_args = (
+                "(rate(%(meter)s{%(label)s='%(label_value)s'}[%(period)ss]))"
+                % {'label': uuid_label_key, 'label_value': instance_label,
+                   'meter': meter, 'period': period}
+            )
+        elif meter == 'libvirt_domain_info_memory_usage_bytes':
+            query_args = (
+                "%(agg)s_over_time(%(meter)s{%(label)s='%(label_value)s'}[%(period)ss]) / 1024"
+                % {'label': uuid_label_key, 'label_value': instance_label,
+                   'meter': meter, 'period': period, 'agg': aggregate}
+            )
         else:
             raise exception.InvalidParameter(
                 message=(_("Cannot process prometheus meter %s") % meter)

watcher/decision_engine/strategy/strategies/host_maintenance.py

@@ -89,6 +89,12 @@ class HostMaintenance(base.HostMaintenanceBaseStrategy):
                                    "will backup the maintenance node.",
                     "type": "string",
                 },
+                "disabled_reason": {
+                        "description": "Describe reason for moving host"
+                                       "into maintenance mode.",
+                        "type": "string",
+                        "default": 'watcher_maintaining',
+                },
                 "disable_live_migration": {
                     "description": "Disable live migration in maintenance. "
                                    "If True, active instances will be cold "
@@ -295,6 +301,10 @@ class HostMaintenance(base.HostMaintenanceBaseStrategy):
 
         maintenance_node = self.input_parameters.get('maintenance_node')
         backup_node = self.input_parameters.get('backup_node')
+        disabled_reason = self.input_parameters.get('disabled_reason')
+        
+        if disabled_reason:
+            self.REASON_FOR_MAINTAINING = disabled_reason
 
         # if no VMs in the maintenance_node, just maintain the compute node
         src_node = self.compute_model.get_node_by_name(maintenance_node)

watcher/decision_engine/strategy/strategies/workload_stabilization.py

@@ -287,8 +287,6 @@ class WorkloadStabilization(base.WorkloadStabilizationBaseStrategy):
                     "for %(metric_name)s", dict(
                         resource_id=instance.uuid, metric_name=meter))
                 return
-            if meter == 'instance_cpu_usage':
-                avg_meter /= float(100)
             LOG.debug('Load of %(metric)s for %(instance)s is %(value)s',
                       {'metric': meter,
                        'instance': instance.uuid,
@@ -334,8 +332,6 @@ class WorkloadStabilization(base.WorkloadStabilizationBaseStrategy):
                 else:
                     if meter_name == 'host_ram_usage':
                         avg_meter /= oslo_utils.units.Ki
-                    if meter_name == 'host_cpu_usage':
-                        avg_meter /= 100
                 LOG.debug('Load of %(metric)s for %(node)s is %(value)s',
                           {'metric': metric,
                            'node': node_id,