osm_policy_module/core/agent.py

   1 # -*- coding: utf-8 -*-
   2
   3 # Copyright 2018 Whitestack, LLC
   4 # *************************************************************
   5
   6 # This file is part of OSM Monitoring module
   7 # All Rights Reserved to Whitestack, LLC
   8
   9 # Licensed under the Apache License, Version 2.0 (the "License"); you may
  10 # not use this file except in compliance with the License. You may obtain
  11 # a copy of the License at
  12
  13 #         http://www.apache.org/licenses/LICENSE-2.0
  14
  15 # Unless required by applicable law or agreed to in writing, software
  16 # distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
  17 # WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
  18 # License for the specific language governing permissions and limitations
  19 # under the License.
  20
  21 # For those usages not covered by the Apache License, Version 2.0 please
  22 # contact: bdiaz@whitestack.com or glavado@whitestack.com
  23 ##
  24 import datetime
  25 import json
  26 import logging
  27 import threading
  28 from json import JSONDecodeError
  29
  30 import yaml
  31 from kafka import KafkaConsumer
  32
  33 from osm_policy_module.common.db_client import DbClient
  34 from osm_policy_module.common.lcm_client import LcmClient
  35 from osm_policy_module.common.mon_client import MonClient
  36 from osm_policy_module.core import database
  37 from osm_policy_module.core.config import Config
  38 from osm_policy_module.core.database import ScalingGroup, ScalingAlarm, ScalingPolicy, ScalingCriteria
  39
  40 log = logging.getLogger(__name__)
  41
  42 ALLOWED_KAFKA_KEYS = ['instantiated', 'scaled', 'notify_alarm']
  43
  44
  45 class PolicyModuleAgent:
  46     def __init__(self):
  47         cfg = Config.instance()
  48         self.db_client = DbClient()
  49         self.mon_client = MonClient()
  50         self.lcm_client = LcmClient()
  51         self.kafka_server = '{}:{}'.format(cfg.OSMPOL_MESSAGE_HOST,
  52                                            cfg.OSMPOL_MESSAGE_PORT)
  53
  54     def run(self):
  55         consumer = KafkaConsumer(bootstrap_servers=self.kafka_server,
  56                                  key_deserializer=bytes.decode,
  57                                  value_deserializer=bytes.decode,
  58                                  group_id='pol-consumer')
  59         consumer.subscribe(["ns", "alarm_response"])
  60
  61         for message in consumer:
  62             t = threading.Thread(target=self._process_msg, args=(message.topic, message.key, message.value,))
  63             t.start()
  64
  65     def _process_msg(self, topic, key, msg):
  66         try:
  67             log.debug("Message arrived with topic: %s, key: %s, msg: %s", topic, key, msg)
  68             if key in ALLOWED_KAFKA_KEYS:
  69                 try:
  70                     content = json.loads(msg)
  71                 except JSONDecodeError:
  72                     content = yaml.safe_load(msg)
  73
  74                 if key == 'instantiated' or key == 'scaled':
  75                     self._handle_instantiated_or_scaled(content)
  76
  77                 if key == 'notify_alarm':
  78                     self._handle_alarm_notification(content)
  79             else:
  80                 log.debug("Key %s is not in ALLOWED_KAFKA_KEYS", key)
  81         except Exception:
  82             log.exception("Error consuming message: ")
  83
  84     def _handle_alarm_notification(self, content):
  85         alarm_id = content['notify_details']['alarm_uuid']
  86         metric_name = content['notify_details']['metric_name']
  87         operation = content['notify_details']['operation']
  88         threshold = content['notify_details']['threshold_value']
  89         vdu_name = content['notify_details']['vdu_name']
  90         vnf_member_index = content['notify_details']['vnf_member_index']
  91         ns_id = content['notify_details']['ns_id']
  92         log.info(
  93             "Received alarm notification for alarm %s, \
  94             metric %s, \
  95             operation %s, \
  96             threshold %s, \
  97             vdu_name %s, \
  98             vnf_member_index %s, \
  99             ns_id %s ",
 100             alarm_id, metric_name, operation, threshold, vdu_name, vnf_member_index, ns_id)
 101         try:
 102             alarm = ScalingAlarm.select().where(ScalingAlarm.alarm_id == alarm_id).get()
 103             delta = datetime.datetime.now() - alarm.scaling_criteria.scaling_policy.last_scale
 104             if delta.total_seconds() < alarm.scaling_criteria.scaling_policy.cooldown_time:
 105                 log.info("Time between last scale and now is less than cooldown time. Skipping.")
 106                 return
 107             log.info("Sending scaling action message for ns: %s", alarm_id)
 108             self.lcm_client.scale(alarm.scaling_criteria.scaling_policy.scaling_group.nsr_id,
 109                                   alarm.scaling_criteria.scaling_policy.scaling_group.name,
 110                                   alarm.vnf_member_index,
 111                                   alarm.action)
 112             alarm.scaling_criteria.scaling_policy.last_scale = datetime.datetime.now()
 113             alarm.scaling_criteria.scaling_policy.save()
 114         except ScalingAlarm.DoesNotExist:
 115             log.info("There is no action configured for alarm %s.", alarm_id)
 116
 117     def _handle_instantiated_or_scaled(self, content):
 118         nslcmop_id = content['nslcmop_id']
 119         nslcmop = self.db_client.get_nslcmop(nslcmop_id)
 120         if nslcmop['operationState'] == 'COMPLETED' or nslcmop['operationState'] == 'PARTIALLY_COMPLETED':
 121             nsr_id = nslcmop['nsInstanceId']
 122             log.info("Configuring scaling groups for network service with nsr_id: %s", nsr_id)
 123             self._configure_scaling_groups(nsr_id)
 124         else:
 125             log.info(
 126                 "Network service is not in COMPLETED or PARTIALLY_COMPLETED state. "
 127                 "Current state is %s. Skipping...",
 128                 nslcmop['operationState'])
 129
 130     def _configure_scaling_groups(self, nsr_id: str):
 131         # TODO(diazb): Check for alarm creation on exception and clean resources if needed.
 132         # TODO: Add support for non-nfvi metrics
 133         with database.db.atomic():
 134             vnfrs = self.db_client.get_vnfrs(nsr_id)
 135             log.info("Checking %s vnfrs...", len(vnfrs))
 136             for vnfr in vnfrs:
 137                 vnfd = self.db_client.get_vnfd(vnfr['vnfd-id'])
 138                 log.info("Looking for vnfd %s", vnfr['vnfd-id'])
 139                 scaling_groups = vnfd['scaling-group-descriptor']
 140                 vnf_monitoring_params = vnfd['monitoring-param']
 141                 for scaling_group in scaling_groups:
 142                     try:
 143                         scaling_group_record = ScalingGroup.select().where(
 144                             ScalingGroup.nsr_id == nsr_id,
 145                             ScalingGroup.name == scaling_group['name']
 146                         ).get()
 147                     except ScalingGroup.DoesNotExist:
 148                         log.info("Creating scaling group record in DB...")
 149                         scaling_group_record = ScalingGroup.create(
 150                             nsr_id=nsr_id,
 151                             name=scaling_group['name'],
 152                             content=json.dumps(scaling_group)
 153                         )
 154                         log.info("Created scaling group record in DB : nsr_id=%s, name=%s, content=%s",
 155                                  scaling_group_record.nsr_id,
 156                                  scaling_group_record.name,
 157                                  scaling_group_record.content)
 158                     for scaling_policy in scaling_group['scaling-policy']:
 159                         if scaling_policy['scaling-type'] != 'automatic':
 160                             continue
 161                         try:
 162                             scaling_policy_record = ScalingPolicy.select().join(ScalingGroup).where(
 163                                 ScalingPolicy.name == scaling_policy['name'],
 164                                 ScalingGroup.id == scaling_group_record.id
 165                             ).get()
 166                         except ScalingPolicy.DoesNotExist:
 167                             log.info("Creating scaling policy record in DB...")
 168                             scaling_policy_record = ScalingPolicy.create(
 169                                 nsr_id=nsr_id,
 170                                 name=scaling_policy['name'],
 171                                 cooldown_time=scaling_policy['cooldown-time'],
 172                                 scaling_group=scaling_group_record
 173                             )
 174                             log.info("Created scaling policy record in DB : name=%s, scaling_group.name=%s",
 175                                      scaling_policy_record.name,
 176                                      scaling_policy_record.scaling_group.name)
 177
 178                         for scaling_criteria in scaling_policy['scaling-criteria']:
 179                             try:
 180                                 scaling_criteria_record = ScalingCriteria.select().join(ScalingPolicy).where(
 181                                     ScalingPolicy.id == scaling_policy_record.id,
 182                                     ScalingCriteria.name == scaling_criteria['name']
 183                                 ).get()
 184                             except ScalingCriteria.DoesNotExist:
 185                                 log.info("Creating scaling criteria record in DB...")
 186                                 scaling_criteria_record = ScalingCriteria.create(
 187                                     nsr_id=nsr_id,
 188                                     name=scaling_criteria['name'],
 189                                     scaling_policy=scaling_policy_record
 190                                 )
 191                                 log.info(
 192                                     "Created scaling criteria record in DB : name=%s, scaling_policy.name=%s",
 193                                     scaling_criteria_record.name,
 194                                     scaling_criteria_record.scaling_policy.name)
 195
 196                             for vdu_ref in scaling_group['vdu']:
 197                                 vnf_monitoring_param = next(
 198                                     filter(lambda param: param['id'] == scaling_criteria['vnf-monitoring-param-ref'],
 199                                            vnf_monitoring_params))
 200                                 if not vdu_ref['vdu-id-ref'] == vnf_monitoring_param['vdu-ref']:
 201                                     continue
 202                                 vdu = next(
 203                                     filter(lambda vdu: vdu['id'] == vdu_ref['vdu-id-ref'], vnfd['vdu'])
 204                                 )
 205                                 vdu_monitoring_params = vdu['monitoring-param']
 206                                 vdu_monitoring_param = next(
 207                                     filter(
 208                                         lambda param: param['id'] == vnf_monitoring_param['vdu-monitoring-param-ref'],
 209                                         vdu_monitoring_params))
 210                                 vdurs = list(filter(lambda vdur: vdur['vdu-id-ref'] == vnf_monitoring_param['vdu-ref'],
 211                                                     vnfr['vdur']))
 212                                 for vdur in vdurs:
 213                                     try:
 214                                         ScalingAlarm.select().join(ScalingCriteria).where(
 215                                             ScalingAlarm.vdu_name == vdur['name'],
 216                                             ScalingCriteria.name == scaling_criteria['name']
 217                                         ).get()
 218                                         log.debug("VDU %s already has an alarm configured", vdur['name'])
 219                                         continue
 220                                     except ScalingAlarm.DoesNotExist:
 221                                         pass
 222                                     alarm_uuid = self.mon_client.create_alarm(
 223                                         metric_name=vdu_monitoring_param['nfvi-metric'],
 224                                         ns_id=nsr_id,
 225                                         vdu_name=vdur['name'],
 226                                         vnf_member_index=vnfr['member-vnf-index-ref'],
 227                                         threshold=scaling_criteria['scale-in-threshold'],
 228                                         operation=scaling_criteria['scale-in-relational-operation'],
 229                                         statistic=vnf_monitoring_param['aggregation-type']
 230                                     )
 231                                     ScalingAlarm.create(
 232                                         alarm_id=alarm_uuid,
 233                                         action='scale_in',
 234                                         vnf_member_index=int(vnfr['member-vnf-index-ref']),
 235                                         vdu_name=vdur['name'],
 236                                         scaling_criteria=scaling_criteria_record
 237                                     )
 238                                     alarm_uuid = self.mon_client.create_alarm(
 239                                         metric_name=vdu_monitoring_param['nfvi-metric'],
 240                                         ns_id=nsr_id,
 241                                         vdu_name=vdur['name'],
 242                                         vnf_member_index=vnfr['member-vnf-index-ref'],
 243                                         threshold=scaling_criteria['scale-out-threshold'],
 244                                         operation=scaling_criteria['scale-out-relational-operation'],
 245                                         statistic=vnf_monitoring_param['aggregation-type']
 246                                     )
 247                                     ScalingAlarm.create(
 248                                         alarm_id=alarm_uuid,
 249                                         action='scale_out',
 250                                         vnf_member_index=int(vnfr['member-vnf-index-ref']),
 251                                         vdu_name=vdur['name'],
 252                                         scaling_criteria=scaling_criteria_record
 253                                     )