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 asyncio
  25 import datetime
  26 import json
  27 import logging
  28 from json import JSONDecodeError
  29
  30 import yaml
  31 from aiokafka import AIOKafkaConsumer
  32
  33 from osm_policy_module.common.common_db_client import CommonDbClient
  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, DatabaseManager
  39
  40 log = logging.getLogger(__name__)
  41
  42 ALLOWED_KAFKA_KEYS = ['instantiated', 'scaled', 'notify_alarm']
  43
  44
  45 class PolicyModuleAgent:
  46     def __init__(self, loop=None):
  47         cfg = Config.instance()
  48         if not loop:
  49             loop = asyncio.get_event_loop()
  50         self.loop = loop
  51         self.db_client = CommonDbClient()
  52         self.mon_client = MonClient(loop=self.loop)
  53         self.lcm_client = LcmClient(loop=self.loop)
  54         self.kafka_server = '{}:{}'.format(cfg.OSMPOL_MESSAGE_HOST,
  55                                            cfg.OSMPOL_MESSAGE_PORT)
  56         self.database_manager = DatabaseManager()
  57
  58     def run(self):
  59         self.loop.run_until_complete(self.start())
  60
  61     async def start(self):
  62         consumer = AIOKafkaConsumer(
  63             "ns",
  64             "alarm_response",
  65             loop=self.loop,
  66             bootstrap_servers=self.kafka_server,
  67             group_id="pol-consumer",
  68             key_deserializer=bytes.decode,
  69             value_deserializer=bytes.decode,
  70         )
  71         await consumer.start()
  72         try:
  73             async for msg in consumer:
  74                 await self._process_msg(msg.topic, msg.key, msg.value)
  75         finally:
  76             await consumer.stop()
  77
  78     async def _process_msg(self, topic, key, msg):
  79         log.debug("_process_msg topic=%s key=%s msg=%s", topic, key, msg)
  80         try:
  81             if key in ALLOWED_KAFKA_KEYS:
  82                 try:
  83                     content = json.loads(msg)
  84                 except JSONDecodeError:
  85                     content = yaml.safe_load(msg)
  86
  87                 if key == 'instantiated' or key == 'scaled':
  88                     await self._handle_instantiated_or_scaled(content)
  89
  90                 if key == 'notify_alarm':
  91                     await self._handle_alarm_notification(content)
  92             else:
  93                 log.debug("Key %s is not in ALLOWED_KAFKA_KEYS", key)
  94         except Exception:
  95             log.exception("Error consuming message: ")
  96
  97     async def _handle_alarm_notification(self, content):
  98         log.debug("_handle_alarm_notification: %s", content)
  99         alarm_id = content['notify_details']['alarm_uuid']
 100         metric_name = content['notify_details']['metric_name']
 101         operation = content['notify_details']['operation']
 102         threshold = content['notify_details']['threshold_value']
 103         vdu_name = content['notify_details']['vdu_name']
 104         vnf_member_index = content['notify_details']['vnf_member_index']
 105         ns_id = content['notify_details']['ns_id']
 106         log.info(
 107             "Received alarm notification for alarm %s, \
 108             metric %s, \
 109             operation %s, \
 110             threshold %s, \
 111             vdu_name %s, \
 112             vnf_member_index %s, \
 113             ns_id %s ",
 114             alarm_id, metric_name, operation, threshold, vdu_name, vnf_member_index, ns_id)
 115         try:
 116             alarm = self.database_manager.get_alarm(alarm_id)
 117             delta = datetime.datetime.now() - alarm.scaling_criteria.scaling_policy.last_scale
 118             log.debug("last_scale: %s", alarm.scaling_criteria.scaling_policy.last_scale)
 119             log.debug("now: %s", datetime.datetime.now())
 120             log.debug("delta: %s", delta)
 121             if delta.total_seconds() < alarm.scaling_criteria.scaling_policy.cooldown_time:
 122                 log.info("Time between last scale and now is less than cooldown time. Skipping.")
 123                 return
 124             log.info("Sending scaling action message for ns: %s", alarm_id)
 125             await self.lcm_client.scale(ns_id,
 126                                         alarm.scaling_criteria.scaling_policy.scaling_group.name,
 127                                         alarm.vnf_member_index,
 128                                         alarm.action)
 129             alarm.scaling_criteria.scaling_policy.last_scale = datetime.datetime.now()
 130             alarm.scaling_criteria.scaling_policy.save()
 131         except ScalingAlarm.DoesNotExist:
 132             log.info("There is no action configured for alarm %s.", alarm_id)
 133
 134     async def _handle_instantiated_or_scaled(self, content):
 135         log.debug("_handle_instantiated_or_scaled: %s", content)
 136         nslcmop_id = content['nslcmop_id']
 137         nslcmop = self.db_client.get_nslcmop(nslcmop_id)
 138         if nslcmop['operationState'] == 'COMPLETED' or nslcmop['operationState'] == 'PARTIALLY_COMPLETED':
 139             nsr_id = nslcmop['nsInstanceId']
 140             log.info("Configuring scaling groups for network service with nsr_id: %s", nsr_id)
 141             await self._configure_scaling_groups(nsr_id)
 142         else:
 143             log.info(
 144                 "Network service is not in COMPLETED or PARTIALLY_COMPLETED state. "
 145                 "Current state is %s. Skipping...",
 146                 nslcmop['operationState'])
 147
 148     async def _configure_scaling_groups(self, nsr_id: str):
 149         log.debug("_configure_scaling_groups: %s", nsr_id)
 150         # TODO: Add support for non-nfvi metrics
 151         alarms_created = []
 152         with database.db.atomic() as tx:
 153             try:
 154                 vnfrs = self.db_client.get_vnfrs(nsr_id)
 155                 for vnfr in vnfrs:
 156                     log.info("Processing vnfr: %s", vnfr)
 157                     vnfd = self.db_client.get_vnfd(vnfr['vnfd-id'])
 158                     log.info("Looking for vnfd %s", vnfr['vnfd-id'])
 159                     if 'scaling-group-descriptor' not in vnfd:
 160                         continue
 161                     scaling_groups = vnfd['scaling-group-descriptor']
 162                     vnf_monitoring_params = vnfd['monitoring-param']
 163                     for scaling_group in scaling_groups:
 164                         try:
 165                             scaling_group_record = ScalingGroup.select().where(
 166                                 ScalingGroup.nsr_id == nsr_id,
 167                                 ScalingGroup.vnf_member_index == int(vnfr['member-vnf-index-ref']),
 168                                 ScalingGroup.name == scaling_group['name']
 169                             ).get()
 170                             log.info("Found existing scaling group record in DB...")
 171                         except ScalingGroup.DoesNotExist:
 172                             log.info("Creating scaling group record in DB...")
 173                             scaling_group_record = ScalingGroup.create(
 174                                 nsr_id=nsr_id,
 175                                 vnf_member_index=vnfr['member-vnf-index-ref'],
 176                                 name=scaling_group['name'],
 177                                 content=json.dumps(scaling_group)
 178                             )
 179                             log.info(
 180                                 "Created scaling group record in DB : nsr_id=%s, vnf_member_index=%s, name=%s",
 181                                 scaling_group_record.nsr_id,
 182                                 scaling_group_record.vnf_member_index,
 183                                 scaling_group_record.name)
 184                         for scaling_policy in scaling_group['scaling-policy']:
 185                             if scaling_policy['scaling-type'] != 'automatic':
 186                                 continue
 187                             try:
 188                                 scaling_policy_record = ScalingPolicy.select().join(ScalingGroup).where(
 189                                     ScalingPolicy.name == scaling_policy['name'],
 190                                     ScalingGroup.id == scaling_group_record.id
 191                                 ).get()
 192                                 log.info("Found existing scaling policy record in DB...")
 193                             except ScalingPolicy.DoesNotExist:
 194                                 log.info("Creating scaling policy record in DB...")
 195                                 scaling_policy_record = ScalingPolicy.create(
 196                                     nsr_id=nsr_id,
 197                                     name=scaling_policy['name'],
 198                                     cooldown_time=scaling_policy['cooldown-time'],
 199                                     scaling_group=scaling_group_record
 200                                 )
 201                                 log.info("Created scaling policy record in DB : name=%s, scaling_group.name=%s",
 202                                          scaling_policy_record.name,
 203                                          scaling_policy_record.scaling_group.name)
 204
 205                             for scaling_criteria in scaling_policy['scaling-criteria']:
 206                                 try:
 207                                     scaling_criteria_record = ScalingCriteria.select().join(ScalingPolicy).where(
 208                                         ScalingPolicy.id == scaling_policy_record.id,
 209                                         ScalingCriteria.name == scaling_criteria['name']
 210                                     ).get()
 211                                     log.info("Found existing scaling criteria record in DB...")
 212                                 except ScalingCriteria.DoesNotExist:
 213                                     log.info("Creating scaling criteria record in DB...")
 214                                     scaling_criteria_record = ScalingCriteria.create(
 215                                         nsr_id=nsr_id,
 216                                         name=scaling_criteria['name'],
 217                                         scaling_policy=scaling_policy_record
 218                                     )
 219                                     log.info(
 220                                         "Created scaling criteria record in DB : name=%s, scaling_policy.name=%s",
 221                                         scaling_criteria_record.name,
 222                                         scaling_criteria_record.scaling_policy.name)
 223
 224                                 vnf_monitoring_param = next(
 225                                     filter(
 226                                         lambda param: param['id'] == scaling_criteria[
 227                                             'vnf-monitoring-param-ref'
 228                                         ],
 229                                         vnf_monitoring_params)
 230                                 )
 231                                 if 'vdu-monitoring-param' in vnf_monitoring_param:
 232                                     vdurs = list(
 233                                         filter(
 234                                             lambda vdur: vdur['vdu-id-ref'] == vnf_monitoring_param
 235                                             ['vdu-monitoring-param']
 236                                             ['vdu-ref'],
 237                                             vnfr['vdur']
 238                                         )
 239                                     )
 240                                 elif 'vdu-metric' in vnf_monitoring_param:
 241                                     vdurs = list(
 242                                         filter(
 243                                             lambda vdur: vdur['vdu-id-ref'] == vnf_monitoring_param
 244                                             ['vdu-metric']
 245                                             ['vdu-ref'],
 246                                             vnfr['vdur']
 247                                         )
 248                                     )
 249                                 elif 'vnf-metric' in vnf_monitoring_param:
 250                                     log.warning("vnf-metric is not currently supported.")
 251                                     continue
 252                                 else:
 253                                     log.warning(
 254                                         "Scaling criteria is referring to a vnf-monitoring-param that does not "
 255                                         "contain a reference to a vdu or vnf metric.")
 256                                     continue
 257                                 for vdur in vdurs:
 258                                     log.info("Creating alarm for vdur %s ", vdur)
 259                                     try:
 260                                         (ScalingAlarm.select()
 261                                          .join(ScalingCriteria)
 262                                          .join(ScalingPolicy)
 263                                          .join(ScalingGroup)
 264                                          .where(
 265                                             ScalingAlarm.vdu_name == vdur['name'],
 266                                             ScalingCriteria.name == scaling_criteria['name'],
 267                                             ScalingPolicy.name == scaling_policy['name'],
 268                                             ScalingGroup.nsr_id == nsr_id
 269                                         ).get())
 270                                         log.debug("vdu %s already has an alarm configured", vdur['name'])
 271                                         continue
 272                                     except ScalingAlarm.DoesNotExist:
 273                                         pass
 274                                     alarm_uuid = await self.mon_client.create_alarm(
 275                                         metric_name=vnf_monitoring_param['id'],
 276                                         ns_id=nsr_id,
 277                                         vdu_name=vdur['name'],
 278                                         vnf_member_index=vnfr['member-vnf-index-ref'],
 279                                         threshold=scaling_criteria['scale-in-threshold'],
 280                                         operation=scaling_criteria['scale-in-relational-operation'],
 281                                         statistic=vnf_monitoring_param['aggregation-type']
 282                                     )
 283                                     ScalingAlarm.create(
 284                                         alarm_id=alarm_uuid,
 285                                         action='scale_in',
 286                                         vnf_member_index=int(vnfr['member-vnf-index-ref']),
 287                                         vdu_name=vdur['name'],
 288                                         scaling_criteria=scaling_criteria_record
 289                                     )
 290                                     alarm_uuid = await self.mon_client.create_alarm(
 291                                         metric_name=vnf_monitoring_param['id'],
 292                                         ns_id=nsr_id,
 293                                         vdu_name=vdur['name'],
 294                                         vnf_member_index=vnfr['member-vnf-index-ref'],
 295                                         threshold=scaling_criteria['scale-out-threshold'],
 296                                         operation=scaling_criteria['scale-out-relational-operation'],
 297                                         statistic=vnf_monitoring_param['aggregation-type']
 298                                     )
 299                                     ScalingAlarm.create(
 300                                         alarm_id=alarm_uuid,
 301                                         action='scale_out',
 302                                         vnf_member_index=int(vnfr['member-vnf-index-ref']),
 303                                         vdu_name=vdur['name'],
 304                                         scaling_criteria=scaling_criteria_record
 305                                     )
 306
 307             except Exception as e:
 308                 log.exception("Error configuring scaling groups:")
 309                 tx.rollback()
 310                 if len(alarms_created) > 0:
 311                     log.info("Cleaning alarm resources in MON")
 312                     for alarm in alarms_created:
 313                         await self.mon_client.delete_alarm(*alarm)
 314                 raise e