osm_policy_module/autoscaling/service.py

   1 # -*- coding: utf-8 -*-
   2 # pylint: disable=no-member
   3
   4 # Copyright 2018 Whitestack, LLC
   5 # *************************************************************
   6
   7 # This file is part of OSM Monitoring module
   8 # All Rights Reserved to Whitestack, LLC
   9
  10 # Licensed under the Apache License, Version 2.0 (the "License"); you may
  11 # not use this file except in compliance with the License. You may obtain
  12 # a copy of the License at
  13
  14 #         http://www.apache.org/licenses/LICENSE-2.0
  15
  16 # Unless required by applicable law or agreed to in writing, software
  17 # distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
  18 # WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
  19 # License for the specific language governing permissions and limitations
  20 # under the License.
  21
  22 # For those usages not covered by the Apache License, Version 2.0 please
  23 # contact: bdiaz@whitestack.com or glavado@whitestack.com
  24 ##
  25 import asyncio
  26 import datetime
  27 import json
  28 import logging
  29
  30 from osm_policy_module.common.common_db_client import CommonDbClient
  31 from osm_policy_module.common.lcm_client import LcmClient
  32 from osm_policy_module.common.mon_client import MonClient
  33 from osm_policy_module.core import database
  34 from osm_policy_module.core.config import Config
  35 from osm_policy_module.core.database import ScalingGroup, ScalingAlarm, ScalingPolicy, ScalingCriteria, \
  36     ScalingAlarmRepository, ScalingGroupRepository, ScalingPolicyRepository, ScalingCriteriaRepository
  37 from osm_policy_module.core.exceptions import VdurNotFound
  38
  39 log = logging.getLogger(__name__)
  40
  41
  42 class AutoscalingService:
  43
  44     def __init__(self, config: Config, loop=None):
  45         self.conf = config
  46         if not loop:
  47             loop = asyncio.get_event_loop()
  48         self.loop = loop
  49         self.db_client = CommonDbClient(config)
  50         self.mon_client = MonClient(config, loop=self.loop)
  51         self.lcm_client = LcmClient(config, loop=self.loop)
  52
  53     async def configure_scaling_groups(self, nsr_id: str):
  54         """
  55         Configures scaling groups for a network service. Creates records in DB. Creates alarms in MON.
  56         :param nsr_id: Network service record id
  57         :return:
  58         """
  59         log.info("Configuring scaling groups for network service with nsr_id: %s",
  60                  nsr_id)
  61         alarms_created = []
  62         database.db.connect()
  63         try:
  64             with database.db.atomic() as tx:
  65                 try:
  66                     vnfrs = self.db_client.get_vnfrs(nsr_id)
  67                     for vnfr in vnfrs:
  68                         log.debug("Processing vnfr: %s", vnfr)
  69                         vnfd = self.db_client.get_vnfd(vnfr['vnfd-id'])
  70                         # TODO: Change for multiple DF support
  71                         df = vnfd.get('df', [{}])[0]
  72                         if 'scaling-aspect' not in df:
  73                             log.debug("No scaling aspect present in vnfd")
  74                             continue
  75                         # TODO: Change for multiple instantiation levels support
  76                         scaling_aspects = df['scaling-aspect']
  77                         all_vnfd_monitoring_params = self._get_all_vnfd_monitoring_params(vnfd)
  78                         for scaling_aspect in scaling_aspects:
  79                             scaling_group_record = self._get_or_create_scaling_group(nsr_id,
  80                                                                                      vnfr['member-vnf-index-ref'],
  81                                                                                      scaling_aspect)
  82                             vdurs = self._get_monitored_vdurs(scaling_aspect, vnfr['vdur'])
  83                             for scaling_policy in scaling_aspect.get('scaling-policy', ()):
  84                                 if scaling_policy['scaling-type'] != 'automatic':
  85                                     continue
  86                                 scaling_policy_record = self._get_or_create_scaling_policy(nsr_id,
  87                                                                                            scaling_policy,
  88                                                                                            scaling_group_record)
  89
  90                                 for scaling_criteria in scaling_policy['scaling-criteria']:
  91                                     scaling_criteria_record = self._get_or_create_scaling_criteria(
  92                                         nsr_id,
  93                                         scaling_criteria,
  94                                         scaling_policy_record
  95                                     )
  96                                     monitoring_param_ref = scaling_criteria.get('vnf-monitoring-param-ref')
  97                                     vnf_monitoring_param = all_vnfd_monitoring_params[monitoring_param_ref]
  98
  99                                     for vdur in vdurs:
 100                                         vdu_id = vdur['vdu-id-ref']
 101                                         log.debug("Creating alarm for vdur %s ", vdur)
 102                                         try:
 103                                             ScalingAlarmRepository.get(ScalingAlarm.vdu_name == vdur['name'],
 104                                                                        ScalingCriteria.name == scaling_criteria['name'],
 105                                                                        ScalingPolicy.name == scaling_policy['name'],
 106                                                                        ScalingGroup.nsr_id == nsr_id,
 107                                                                        join_classes=[ScalingCriteria,
 108                                                                                      ScalingPolicy,
 109                                                                                      ScalingGroup])
 110                                             log.debug("vdu %s already has an alarm configured", vdur['name'])
 111                                             continue
 112                                         except ScalingAlarm.DoesNotExist:
 113                                             pass
 114                                         metric_name = self._get_metric_name(vnf_monitoring_param)
 115
 116                                         db_nsr = self.db_client.get_nsr(nsr_id)
 117                                         nb_scale_op = 0
 118                                         if db_nsr["_admin"].get("scaling-group"):
 119                                             db_nsr_admin = db_nsr["_admin"]["scaling-group"]
 120                                             for admin_scale_index, admin_scale_info in enumerate(db_nsr_admin):
 121                                                 if admin_scale_info["name"] == scaling_aspect["name"]:
 122                                                     nb_scale_op = admin_scale_info.get("nb-scale-op", 0)
 123                                                     break
 124                                         min_instance_count = 1
 125                                         for vdu_profile in df.get('vdu-profile', ()):
 126                                             if vdu_profile.get('id') == vdu_id:
 127                                                 min_instance_count = int(vdu_profile.get('min-number-of-instances ', 1))
 128                                                 break
 129                                         if nb_scale_op >= min_instance_count:
 130                                             alarm_uuid = await self.mon_client.create_alarm(
 131                                                 metric_name=metric_name,
 132                                                 ns_id=nsr_id,
 133                                                 vdu_name=vdur['name'],
 134                                                 vnf_member_index=vnfr['member-vnf-index-ref'],
 135                                                 threshold=scaling_criteria['scale-in-threshold'],
 136                                                 operation=scaling_criteria['scale-in-relational-operation']
 137                                             )
 138                                             alarm = ScalingAlarmRepository.create(
 139                                                 alarm_uuid=alarm_uuid,
 140                                                 action='scale_in',
 141                                                 vnf_member_index=vnfr['member-vnf-index-ref'],
 142                                                 vdu_name=vdur['name'],
 143                                                 scaling_criteria=scaling_criteria_record
 144                                             )
 145                                             alarms_created.append(alarm)
 146                                         alarm_uuid = await self.mon_client.create_alarm(
 147                                             metric_name=metric_name,
 148                                             ns_id=nsr_id,
 149                                             vdu_name=vdur['name'],
 150                                             vnf_member_index=vnfr['member-vnf-index-ref'],
 151                                             threshold=scaling_criteria['scale-out-threshold'],
 152                                             operation=scaling_criteria['scale-out-relational-operation']
 153                                         )
 154                                         alarm = ScalingAlarmRepository.create(
 155                                             alarm_uuid=alarm_uuid,
 156                                             action='scale_out',
 157                                             vnf_member_index=vnfr['member-vnf-index-ref'],
 158                                             vdu_name=vdur['name'],
 159                                             scaling_criteria=scaling_criteria_record
 160                                         )
 161                                         alarms_created.append(alarm)
 162
 163                 except Exception as e:
 164                     log.exception("Error configuring scaling groups:")
 165                     tx.rollback()
 166                     if len(alarms_created) > 0:
 167                         log.info("Cleaning alarm resources in MON")
 168                         for alarm in alarms_created:
 169                             await self.mon_client.delete_alarm(
 170                                 alarm.scaling_criteria.scaling_policy.scaling_group.nsr_id,
 171                                 alarm.vnf_member_index,
 172                                 alarm.vdu_name,
 173                                 alarm.alarm_uuid)
 174                     raise e
 175         finally:
 176             database.db.close()
 177
 178     async def delete_scaling_groups(self, nsr_id: str):
 179         log.debug("Deleting scaling groups for network service %s", nsr_id)
 180         database.db.connect()
 181         try:
 182             with database.db.atomic() as tx:
 183                 try:
 184                     for scaling_group in ScalingGroupRepository.list(ScalingGroup.nsr_id == nsr_id):
 185                         for scaling_policy in scaling_group.scaling_policies:
 186                             for scaling_criteria in scaling_policy.scaling_criterias:
 187                                 for alarm in scaling_criteria.scaling_alarms:
 188                                     try:
 189                                         await self.mon_client.delete_alarm(
 190                                             alarm.scaling_criteria.scaling_policy.scaling_group.nsr_id,
 191                                             alarm.vnf_member_index,
 192                                             alarm.vdu_name,
 193                                             alarm.alarm_uuid)
 194                                     except ValueError:
 195                                         log.exception("Error deleting alarm in MON %s", alarm.alarm_uuid)
 196                                     alarm.delete_instance()
 197                                 scaling_criteria.delete_instance()
 198                             scaling_policy.delete_instance()
 199                         scaling_group.delete_instance()
 200
 201                 except Exception as e:
 202                     log.exception("Error deleting scaling groups and alarms:")
 203                     tx.rollback()
 204                     raise e
 205         finally:
 206             database.db.close()
 207
 208     async def delete_orphaned_alarms(self, nsr_id):
 209         log.info("Deleting orphaned scaling alarms for network service %s", nsr_id)
 210         database.db.connect()
 211         try:
 212             with database.db.atomic() as tx:
 213                 try:
 214                     for scaling_group in ScalingGroupRepository.list(ScalingGroup.nsr_id == nsr_id):
 215                         for scaling_policy in scaling_group.scaling_policies:
 216                             for scaling_criteria in scaling_policy.scaling_criterias:
 217                                 for alarm in scaling_criteria.scaling_alarms:
 218                                     try:
 219                                         self.db_client.get_vdur(nsr_id, alarm.vnf_member_index, alarm.vdu_name)
 220                                     except VdurNotFound:
 221                                         log.debug("Deleting orphaned scaling alarm %s", alarm.alarm_uuid)
 222                                         try:
 223                                             await self.mon_client.delete_alarm(
 224                                                 alarm.scaling_criteria.scaling_policy.scaling_group.nsr_id,
 225                                                 alarm.vnf_member_index,
 226                                                 alarm.vdu_name,
 227                                                 alarm.alarm_uuid)
 228                                         except ValueError:
 229                                             log.exception("Error deleting alarm in MON %s", alarm.alarm_uuid)
 230                                         alarm.delete_instance()
 231
 232                 except Exception as e:
 233                     log.exception("Error deleting orphaned alarms:")
 234                     tx.rollback()
 235                     raise e
 236         finally:
 237             database.db.close()
 238
 239     async def handle_alarm(self, alarm_uuid: str, status: str):
 240         await self.update_alarm_status(alarm_uuid, status)
 241         await self.evaluate_policy(alarm_uuid)
 242
 243     async def update_alarm_status(self, alarm_uuid: str, status: str):
 244         database.db.connect()
 245         try:
 246             with database.db.atomic():
 247                 alarm = ScalingAlarmRepository.get(ScalingAlarm.alarm_uuid == alarm_uuid)
 248                 alarm.last_status = status
 249                 alarm.save()
 250         except ScalingAlarm.DoesNotExist:
 251             log.debug("There is no autoscaling action configured for alarm %s.", alarm_uuid)
 252         finally:
 253             database.db.close()
 254
 255     async def evaluate_policy(self, alarm_uuid):
 256         database.db.connect()
 257         try:
 258             with database.db.atomic():
 259                 alarm = ScalingAlarmRepository.get(ScalingAlarm.alarm_uuid == alarm_uuid)
 260                 vnf_member_index = alarm.vnf_member_index
 261                 action = alarm.action
 262                 scaling_policy = alarm.scaling_criteria.scaling_policy
 263                 if not scaling_policy.enabled:
 264                     return
 265                 if action == 'scale_in':
 266                     operation = scaling_policy.scale_in_operation
 267                 elif action == 'scale_out':
 268                     operation = scaling_policy.scale_out_operation
 269                 else:
 270                     raise Exception('Unknown alarm action {}'.format(alarm.action))
 271                 alarms = ScalingAlarmRepository.list(ScalingAlarm.scaling_criteria == alarm.scaling_criteria,
 272                                                      ScalingAlarm.action == alarm.action,
 273                                                      ScalingAlarm.vnf_member_index == vnf_member_index,
 274                                                      ScalingAlarm.vdu_name == alarm.vdu_name)
 275                 statuses = []
 276                 for alarm in alarms:
 277                     statuses.append(alarm.last_status)
 278                 if (operation == 'AND' and set(statuses) == {'alarm'}) or (operation == 'OR' and 'alarm' in statuses):
 279                     delta = datetime.datetime.now() - scaling_policy.last_scale
 280                     if delta.total_seconds() > scaling_policy.cooldown_time:
 281                         log.info("Sending %s action message for ns: %s",
 282                                  alarm.action,
 283                                  scaling_policy.scaling_group.nsr_id)
 284                         await self.lcm_client.scale(scaling_policy.scaling_group.nsr_id,
 285                                                     scaling_policy.scaling_group.name,
 286                                                     vnf_member_index,
 287                                                     action)
 288                         scaling_policy.last_scale = datetime.datetime.now()
 289                         scaling_policy.save()
 290
 291         except ScalingAlarm.DoesNotExist:
 292             log.debug("There is no autoscaling action configured for alarm %s.", alarm_uuid)
 293         finally:
 294             database.db.close()
 295
 296     def _get_all_vnfd_monitoring_params(self, vnfd):
 297         all_monitoring_params = {}
 298         for ivld in vnfd.get("int-virtual-link-desc", ()):
 299             for mp in ivld.get("monitoring-parameters", ()):
 300                 all_monitoring_params[mp.get("id")] = mp
 301
 302         for vdu in vnfd.get("vdu", ()):
 303             for mp in vdu.get("monitoring-parameter", ()):
 304                 all_monitoring_params[mp.get("id")] = mp
 305
 306         for df in vnfd.get("df", ()):
 307             for mp in df.get("monitoring-parameter", ()):
 308                 all_monitoring_params[mp.get("id")] = mp
 309
 310         return all_monitoring_params
 311
 312     def _get_or_create_scaling_group(self, nsr_id: str, vnf_member_index: str, scaling_aspect: dict):
 313         try:
 314             scaling_group_record = ScalingGroupRepository.get(
 315                 ScalingGroup.nsr_id == nsr_id,
 316                 ScalingGroup.vnf_member_index == vnf_member_index,
 317                 ScalingGroup.name == scaling_aspect['name']
 318             )
 319             log.debug("Found existing scaling group record in DB...")
 320         except ScalingGroup.DoesNotExist:
 321             log.debug("Creating scaling group record in DB...")
 322             scaling_group_record = ScalingGroupRepository.create(
 323                 nsr_id=nsr_id,
 324                 vnf_member_index=vnf_member_index,
 325                 name=scaling_aspect['name'],
 326                 content=json.dumps(scaling_aspect)
 327             )
 328             log.debug(
 329                 "Created scaling group record in DB : nsr_id=%s, vnf_member_index=%s, name=%s",
 330                 scaling_group_record.nsr_id,
 331                 scaling_group_record.vnf_member_index,
 332                 scaling_group_record.name)
 333         return scaling_group_record
 334
 335     def _get_or_create_scaling_policy(self, nsr_id: str, scaling_policy: dict, scaling_group_record: ScalingGroup):
 336         try:
 337             scaling_policy_record = ScalingPolicyRepository.get(
 338                 ScalingPolicy.name == scaling_policy['name'],
 339                 ScalingGroup.id == scaling_group_record.id,
 340                 join_classes=[ScalingGroup]
 341             )
 342             log.debug("Found existing scaling policy record in DB...")
 343         except ScalingPolicy.DoesNotExist:
 344             log.debug("Creating scaling policy record in DB...")
 345             scaling_policy_record = ScalingPolicyRepository.create(
 346                 nsr_id=nsr_id,
 347                 name=scaling_policy['name'],
 348                 cooldown_time=scaling_policy['cooldown-time'],
 349                 scaling_group=scaling_group_record,
 350             )
 351             if 'scale-in-operation-type' in scaling_policy:
 352                 scaling_policy_record.scale_in_operation = scaling_policy['scale-in-operation-type']
 353             if 'scale-out-operation-type' in scaling_policy:
 354                 scaling_policy_record.scale_out_operation = scaling_policy['scale-out-operation-type']
 355             if 'enabled' in scaling_policy:
 356                 scaling_policy_record.enabled = scaling_policy['enabled']
 357             scaling_policy_record.save()
 358             log.debug("Created scaling policy record in DB : name=%s, scaling_group.name=%s",
 359                       scaling_policy_record.name,
 360                       scaling_policy_record.scaling_group.name)
 361         return scaling_policy_record
 362
 363     def _get_or_create_scaling_criteria(self, nsr_id: str, scaling_criteria: dict,
 364                                         scaling_policy_record: ScalingPolicy):
 365         try:
 366             scaling_criteria_record = ScalingCriteriaRepository.get(
 367                 ScalingPolicy.id == scaling_policy_record.id,
 368                 ScalingCriteria.name == scaling_criteria['name'],
 369                 join_classes=[ScalingPolicy]
 370             )
 371             log.debug("Found existing scaling criteria record in DB...")
 372         except ScalingCriteria.DoesNotExist:
 373             log.debug("Creating scaling criteria record in DB...")
 374             scaling_criteria_record = ScalingCriteriaRepository.create(
 375                 nsr_id=nsr_id,
 376                 name=scaling_criteria['name'],
 377                 scaling_policy=scaling_policy_record
 378             )
 379             log.debug(
 380                 "Created scaling criteria record in DB : name=%s, scaling_policy.name=%s",
 381                 scaling_criteria_record.name,
 382                 scaling_criteria_record.scaling_policy.name)
 383         return scaling_criteria_record
 384
 385     def _get_monitored_vdurs(self, scaling_aspect: dict, vdurs):
 386         all_monitored_vdus = set()
 387         for delta in scaling_aspect.get('aspect-delta-details', {}).get('deltas', ()):
 388             for vdu_delta in delta.get('vdu-delta', ()):
 389                 all_monitored_vdus.add(vdu_delta.get('id'))
 390
 391         monitored_vdurs = list(filter(lambda vdur: vdur['vdu-id-ref'] in all_monitored_vdus, vdurs))
 392
 393         if not monitored_vdurs:
 394             log.warning(
 395                 "Scaling criteria is referring to a vnf-monitoring-param that does not "
 396                 "contain a reference to a vdu or vnf metric.")
 397         return monitored_vdurs
 398
 399     def _get_metric_name(self, vnf_monitoring_param: dict):
 400         if 'performance-metric' in vnf_monitoring_param:
 401             return vnf_monitoring_param['performance-metric']
 402         raise ValueError('No metric name found for vnf_monitoring_param %s' % vnf_monitoring_param['id'])