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