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