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 import operator
  30 import functools
  31
  32 from osm_policy_module.common.common_db_client import CommonDbClient
  33 from osm_policy_module.common.lcm_client import LcmClient
  34 from osm_policy_module.common.mon_client import MonClient
  35 from osm_policy_module.core import database
  36 from osm_policy_module.core.config import Config
  37 from osm_policy_module.core.database import (
  38     ScalingGroup,
  39     ScalingAlarm,
  40     ScalingPolicy,
  41     ScalingCriteria,
  42     ScalingAlarmRepository,
  43     ScalingGroupRepository,
  44     ScalingPolicyRepository,
  45     ScalingCriteriaRepository,
  46 )
  47 from osm_policy_module.core.exceptions import VdurNotFound
  48
  49 log = logging.getLogger(__name__)
  50
  51
  52 class AutoscalingService:
  53     def __init__(self, config: Config, loop=None):
  54         self.conf = config
  55         if not loop:
  56             loop = asyncio.get_event_loop()
  57         self.loop = loop
  58         self.db_client = CommonDbClient(config)
  59         self.mon_client = MonClient(config, loop=self.loop)
  60         self.lcm_client = LcmClient(config, loop=self.loop)
  61
  62     async def configure_scaling_groups(self, nsr_id: str):
  63         """
  64         Configures scaling groups for a network service. Creates records in DB. Creates alarms in MON.
  65         :param nsr_id: Network service record id
  66         :return:
  67         """
  68         log.info(
  69             "Configuring scaling groups for network service with nsr_id: %s", nsr_id
  70         )
  71         alarms_created = []
  72         database.db.connect()
  73         try:
  74             with database.db.atomic() as tx:
  75                 try:
  76                     vnfrs = self.db_client.get_vnfrs(nsr_id)
  77                     for vnfr in vnfrs:
  78                         log.debug("Processing vnfr: %s", vnfr)
  79                         vnfd = self.db_client.get_vnfd(vnfr["vnfd-id"])
  80                         # TODO: Change for multiple DF support
  81                         df = vnfd.get("df", [{}])[0]
  82                         if "scaling-aspect" not in df:
  83                             log.debug("No scaling aspect present in vnfd")
  84                             continue
  85                         # TODO: Change for multiple instantiation levels support
  86                         scaling_aspects = df["scaling-aspect"]
  87                         all_vnfd_monitoring_params = (
  88                             self._get_all_vnfd_monitoring_params(vnfd)
  89                         )
  90                         for scaling_aspect in scaling_aspects:
  91                             scaling_group_record = self._get_or_create_scaling_group(
  92                                 nsr_id, vnfr["member-vnf-index-ref"], scaling_aspect
  93                             )
  94                             vdurs = self._get_monitored_vdurs(
  95                                 scaling_aspect, vnfr["vdur"]
  96                             )
  97                             for scaling_policy in scaling_aspect.get(
  98                                 "scaling-policy", ()
  99                             ):
 100                                 if scaling_policy["scaling-type"] != "automatic":
 101                                     continue
 102                                 scaling_policy_record = (
 103                                     self._get_or_create_scaling_policy(
 104                                         nsr_id, scaling_policy, scaling_group_record
 105                                     )
 106                                 )
 107
 108                                 for scaling_criteria in scaling_policy[
 109                                     "scaling-criteria"
 110                                 ]:
 111                                     scaling_criteria_record = (
 112                                         self._get_or_create_scaling_criteria(
 113                                             nsr_id,
 114                                             scaling_criteria,
 115                                             scaling_policy_record,
 116                                         )
 117                                     )
 118                                     monitoring_param_ref = scaling_criteria.get(
 119                                         "vnf-monitoring-param-ref"
 120                                     )
 121                                     vnf_monitoring_param = all_vnfd_monitoring_params[
 122                                         monitoring_param_ref
 123                                     ]
 124
 125                                     for vdur in vdurs:
 126                                         vdu_id = vdur["vdu-id-ref"]
 127                                         log.debug("Creating alarm for vdur %s ", vdur)
 128                                         try:
 129                                             ScalingAlarmRepository.get(
 130                                                 ScalingAlarm.vdu_name == vdur["name"],
 131                                                 ScalingCriteria.name
 132                                                 == scaling_criteria["name"],
 133                                                 ScalingPolicy.name
 134                                                 == scaling_policy["name"],
 135                                                 ScalingGroup.nsr_id == nsr_id,
 136                                                 join_classes=[
 137                                                     ScalingCriteria,
 138                                                     ScalingPolicy,
 139                                                     ScalingGroup,
 140                                                 ],
 141                                             )
 142                                             log.debug(
 143                                                 "vdu %s already has an alarm configured",
 144                                                 vdur["name"],
 145                                             )
 146                                             continue
 147                                         except ScalingAlarm.DoesNotExist:
 148                                             pass
 149                                         metric_name = self._get_metric_name(
 150                                             vnf_monitoring_param
 151                                         )
 152
 153                                         db_nsr = self.db_client.get_nsr(nsr_id)
 154                                         nb_scale_op = 0
 155                                         if db_nsr["_admin"].get("scaling-group"):
 156                                             db_nsr_admin = db_nsr["_admin"][
 157                                                 "scaling-group"
 158                                             ]
 159                                             for (
 160                                                 admin_scale_index,
 161                                                 admin_scale_info,
 162                                             ) in enumerate(db_nsr_admin):
 163                                                 if (
 164                                                     admin_scale_info["name"]
 165                                                     == scaling_aspect["name"]
 166                                                 ):
 167                                                     nb_scale_op = admin_scale_info.get(
 168                                                         "nb-scale-op", 0
 169                                                     )
 170                                                     break
 171                                         min_instance_count = 1
 172                                         for vdu_profile in df.get("vdu-profile", ()):
 173                                             if vdu_profile.get("id") == vdu_id:
 174                                                 min_instance_count = int(
 175                                                     vdu_profile.get(
 176                                                         "min-number-of-instances ", 1
 177                                                     )
 178                                                 )
 179                                                 break
 180                                         if nb_scale_op >= min_instance_count:
 181                                             alarm_uuid = (
 182                                                 await self.mon_client.create_alarm(
 183                                                     metric_name=metric_name,
 184                                                     ns_id=nsr_id,
 185                                                     vdu_name=vdur["name"],
 186                                                     vnf_member_index=vnfr[
 187                                                         "member-vnf-index-ref"
 188                                                     ],
 189                                                     threshold=scaling_criteria[
 190                                                         "scale-in-threshold"
 191                                                     ],
 192                                                     operation=scaling_criteria[
 193                                                         "scale-in-relational-operation"
 194                                                     ],
 195                                                     action="scale_in"
 196                                                 )
 197                                             )
 198                                             alarm = ScalingAlarmRepository.create(
 199                                                 alarm_uuid=alarm_uuid,
 200                                                 action="scale_in",
 201                                                 vnf_member_index=vnfr[
 202                                                     "member-vnf-index-ref"
 203                                                 ],
 204                                                 vdu_name=vdur["name"],
 205                                                 scaling_criteria=scaling_criteria_record,
 206                                             )
 207                                             alarms_created.append(alarm)
 208                                         alarm_uuid = await self.mon_client.create_alarm(
 209                                             metric_name=metric_name,
 210                                             ns_id=nsr_id,
 211                                             vdu_name=vdur["name"],
 212                                             vnf_member_index=vnfr[
 213                                                 "member-vnf-index-ref"
 214                                             ],
 215                                             threshold=scaling_criteria[
 216                                                 "scale-out-threshold"
 217                                             ],
 218                                             operation=scaling_criteria[
 219                                                 "scale-out-relational-operation"
 220                                             ],
 221                                             action="scale_out",
 222                                         )
 223                                         alarm = ScalingAlarmRepository.create(
 224                                             alarm_uuid=alarm_uuid,
 225                                             action="scale_out",
 226                                             vnf_member_index=vnfr[
 227                                                 "member-vnf-index-ref"
 228                                             ],
 229                                             vdu_name=vdur["name"],
 230                                             scaling_criteria=scaling_criteria_record,
 231                                         )
 232                                         alarms_created.append(alarm)
 233
 234                 except Exception as e:
 235                     log.exception("Error configuring scaling groups:")
 236                     tx.rollback()
 237                     if len(alarms_created) > 0:
 238                         log.info("Cleaning alarm resources in MON")
 239                         for alarm in alarms_created:
 240                             await self.mon_client.delete_alarm(
 241                                 alarm.scaling_criteria.scaling_policy.scaling_group.nsr_id,
 242                                 alarm.vnf_member_index,
 243                                 alarm.vdu_name,
 244                                 alarm.alarm_uuid,
 245                             )
 246                     raise e
 247         finally:
 248             database.db.close()
 249
 250     async def delete_scaling_groups(self, nsr_id: str, vnf_member_index=None):
 251         log.debug("Deleting scaling groups for network service %s", nsr_id)
 252         database.db.connect()
 253         try:
 254             with database.db.atomic() as tx:
 255                 try:
 256                     if vnf_member_index is None:
 257                         scale_conditions = ScalingGroup.nsr_id == nsr_id
 258                     else:
 259                         query_list = [ScalingGroup.nsr_id == nsr_id,
 260                                       ScalingGroup.vnf_member_index == vnf_member_index]
 261                         scale_conditions = functools.reduce(operator.and_, query_list)
 262                     for scaling_group in ScalingGroupRepository.list(scale_conditions):
 263                         for scaling_policy in scaling_group.scaling_policies:
 264                             for scaling_criteria in scaling_policy.scaling_criterias:
 265                                 for alarm in scaling_criteria.scaling_alarms:
 266                                     try:
 267                                         await self.mon_client.delete_alarm(
 268                                             alarm.scaling_criteria.scaling_policy.scaling_group.nsr_id,
 269                                             alarm.vnf_member_index,
 270                                             alarm.vdu_name,
 271                                             alarm.alarm_uuid,
 272                                         )
 273                                     except ValueError:
 274                                         log.exception(
 275                                             "Error deleting alarm in MON %s",
 276                                             alarm.alarm_uuid,
 277                                         )
 278                                     alarm.delete_instance()
 279                                 scaling_criteria.delete_instance()
 280                             scaling_policy.delete_instance()
 281                         scaling_group.delete_instance()
 282
 283                 except Exception as e:
 284                     log.exception("Error deleting scaling groups and alarms:")
 285                     tx.rollback()
 286                     raise e
 287         finally:
 288             database.db.close()
 289
 290     async def delete_orphaned_alarms(self, nsr_id):
 291         log.info("Deleting orphaned scaling alarms for network service %s", nsr_id)
 292         database.db.connect()
 293         try:
 294             with database.db.atomic() as tx:
 295                 try:
 296                     for scaling_group in ScalingGroupRepository.list(
 297                         ScalingGroup.nsr_id == nsr_id
 298                     ):
 299                         for scaling_policy in scaling_group.scaling_policies:
 300                             for scaling_criteria in scaling_policy.scaling_criterias:
 301                                 for alarm in scaling_criteria.scaling_alarms:
 302                                     try:
 303                                         self.db_client.get_vdur(
 304                                             nsr_id,
 305                                             alarm.vnf_member_index,
 306                                             alarm.vdu_name,
 307                                         )
 308                                     except VdurNotFound:
 309                                         log.debug(
 310                                             "Deleting orphaned scaling alarm %s",
 311                                             alarm.alarm_uuid,
 312                                         )
 313                                         try:
 314                                             await self.mon_client.delete_alarm(
 315                                                 alarm.scaling_criteria.scaling_policy.scaling_group.nsr_id,
 316                                                 alarm.vnf_member_index,
 317                                                 alarm.vdu_name,
 318                                                 alarm.alarm_uuid,
 319                                             )
 320                                         except ValueError:
 321                                             log.exception(
 322                                                 "Error deleting alarm in MON %s",
 323                                                 alarm.alarm_uuid,
 324                                             )
 325                                         alarm.delete_instance()
 326
 327                 except Exception as e:
 328                     log.exception("Error deleting orphaned alarms:")
 329                     tx.rollback()
 330                     raise e
 331         finally:
 332             database.db.close()
 333
 334     async def handle_alarm(self, alarm_uuid: str, status: str):
 335         await self.update_alarm_status(alarm_uuid, status)
 336         await self.evaluate_policy(alarm_uuid)
 337
 338     async def update_alarm_status(self, alarm_uuid: str, status: str):
 339         database.db.connect()
 340         try:
 341             with database.db.atomic():
 342                 alarm = ScalingAlarmRepository.get(
 343                     ScalingAlarm.alarm_uuid == alarm_uuid
 344                 )
 345                 alarm.last_status = status
 346                 alarm.save()
 347         except ScalingAlarm.DoesNotExist:
 348             log.debug(
 349                 "There is no autoscaling action configured for alarm %s.", alarm_uuid
 350             )
 351         finally:
 352             database.db.close()
 353
 354     async def evaluate_policy(self, alarm_uuid):
 355         database.db.connect()
 356         try:
 357             with database.db.atomic():
 358                 alarm = ScalingAlarmRepository.get(
 359                     ScalingAlarm.alarm_uuid == alarm_uuid
 360                 )
 361                 vnf_member_index = alarm.vnf_member_index
 362                 action = alarm.action
 363                 scaling_policy = alarm.scaling_criteria.scaling_policy
 364                 if not scaling_policy.enabled:
 365                     return
 366                 if action == "scale_in":
 367                     operation = scaling_policy.scale_in_operation
 368                 elif action == "scale_out":
 369                     operation = scaling_policy.scale_out_operation
 370                 else:
 371                     raise Exception("Unknown alarm action {}".format(alarm.action))
 372                 alarms = ScalingAlarmRepository.list(
 373                     ScalingAlarm.scaling_criteria == alarm.scaling_criteria,
 374                     ScalingAlarm.action == alarm.action,
 375                     ScalingAlarm.vnf_member_index == vnf_member_index,
 376                     ScalingAlarm.vdu_name == alarm.vdu_name,
 377                 )
 378                 statuses = []
 379                 for alarm in alarms:
 380                     statuses.append(alarm.last_status)
 381                 if (operation == "AND" and set(statuses) == {"alarm"}) or (
 382                     operation == "OR" and "alarm" in statuses
 383                 ):
 384                     delta = datetime.datetime.now() - scaling_policy.last_scale
 385                     if delta.total_seconds() > scaling_policy.cooldown_time:
 386                         log.info(
 387                             "Sending %s action message for ns: %s",
 388                             alarm.action,
 389                             scaling_policy.scaling_group.nsr_id,
 390                         )
 391                         await self.lcm_client.scale(
 392                             scaling_policy.scaling_group.nsr_id,
 393                             scaling_policy.scaling_group.name,
 394                             vnf_member_index,
 395                             action,
 396                         )
 397                         scaling_policy.last_scale = datetime.datetime.now()
 398                         scaling_policy.save()
 399
 400         except ScalingAlarm.DoesNotExist:
 401             log.debug(
 402                 "There is no autoscaling action configured for alarm %s.", alarm_uuid
 403             )
 404         finally:
 405             database.db.close()
 406
 407     def _get_all_vnfd_monitoring_params(self, vnfd):
 408         all_monitoring_params = {}
 409         for ivld in vnfd.get("int-virtual-link-desc", ()):
 410             for mp in ivld.get("monitoring-parameters", ()):
 411                 all_monitoring_params[mp.get("id")] = mp
 412
 413         for vdu in vnfd.get("vdu", ()):
 414             for mp in vdu.get("monitoring-parameter", ()):
 415                 all_monitoring_params[mp.get("id")] = mp
 416
 417         for df in vnfd.get("df", ()):
 418             for mp in df.get("monitoring-parameter", ()):
 419                 all_monitoring_params[mp.get("id")] = mp
 420
 421         return all_monitoring_params
 422
 423     def _get_or_create_scaling_group(
 424         self, nsr_id: str, vnf_member_index: str, scaling_aspect: dict
 425     ):
 426         try:
 427             scaling_group_record = ScalingGroupRepository.get(
 428                 ScalingGroup.nsr_id == nsr_id,
 429                 ScalingGroup.vnf_member_index == vnf_member_index,
 430                 ScalingGroup.name == scaling_aspect["name"],
 431             )
 432             log.debug("Found existing scaling group record in DB...")
 433         except ScalingGroup.DoesNotExist:
 434             log.debug("Creating scaling group record in DB...")
 435             scaling_group_record = ScalingGroupRepository.create(
 436                 nsr_id=nsr_id,
 437                 vnf_member_index=vnf_member_index,
 438                 name=scaling_aspect["name"],
 439                 content=json.dumps(scaling_aspect),
 440             )
 441             log.debug(
 442                 "Created scaling group record in DB : nsr_id=%s, vnf_member_index=%s, name=%s",
 443                 scaling_group_record.nsr_id,
 444                 scaling_group_record.vnf_member_index,
 445                 scaling_group_record.name,
 446             )
 447         return scaling_group_record
 448
 449     def _get_or_create_scaling_policy(
 450         self, nsr_id: str, scaling_policy: dict, scaling_group_record: ScalingGroup
 451     ):
 452         try:
 453             scaling_policy_record = ScalingPolicyRepository.get(
 454                 ScalingPolicy.name == scaling_policy["name"],
 455                 ScalingGroup.id == scaling_group_record.id,
 456                 join_classes=[ScalingGroup],
 457             )
 458             log.debug("Found existing scaling policy record in DB...")
 459         except ScalingPolicy.DoesNotExist:
 460             log.debug("Creating scaling policy record in DB...")
 461             scaling_policy_record = ScalingPolicyRepository.create(
 462                 nsr_id=nsr_id,
 463                 name=scaling_policy["name"],
 464                 cooldown_time=scaling_policy["cooldown-time"],
 465                 scaling_group=scaling_group_record,
 466             )
 467             if "scale-in-operation-type" in scaling_policy:
 468                 scaling_policy_record.scale_in_operation = scaling_policy[
 469                     "scale-in-operation-type"
 470                 ]
 471             if "scale-out-operation-type" in scaling_policy:
 472                 scaling_policy_record.scale_out_operation = scaling_policy[
 473                     "scale-out-operation-type"
 474                 ]
 475             if "enabled" in scaling_policy:
 476                 scaling_policy_record.enabled = scaling_policy["enabled"]
 477             scaling_policy_record.save()
 478             log.debug(
 479                 "Created scaling policy record in DB : name=%s, scaling_group.name=%s",
 480                 scaling_policy_record.name,
 481                 scaling_policy_record.scaling_group.name,
 482             )
 483         return scaling_policy_record
 484
 485     def _get_or_create_scaling_criteria(
 486         self, nsr_id: str, scaling_criteria: dict, scaling_policy_record: ScalingPolicy
 487     ):
 488         try:
 489             scaling_criteria_record = ScalingCriteriaRepository.get(
 490                 ScalingPolicy.id == scaling_policy_record.id,
 491                 ScalingCriteria.name == scaling_criteria["name"],
 492                 join_classes=[ScalingPolicy],
 493             )
 494             log.debug("Found existing scaling criteria record in DB...")
 495         except ScalingCriteria.DoesNotExist:
 496             log.debug("Creating scaling criteria record in DB...")
 497             scaling_criteria_record = ScalingCriteriaRepository.create(
 498                 nsr_id=nsr_id,
 499                 name=scaling_criteria["name"],
 500                 scaling_policy=scaling_policy_record,
 501             )
 502             log.debug(
 503                 "Created scaling criteria record in DB : name=%s, scaling_policy.name=%s",
 504                 scaling_criteria_record.name,
 505                 scaling_criteria_record.scaling_policy.name,
 506             )
 507         return scaling_criteria_record
 508
 509     def _get_monitored_vdurs(self, scaling_aspect: dict, vdurs):
 510         all_monitored_vdus = set()
 511         for delta in scaling_aspect.get("aspect-delta-details", {}).get("deltas", ()):
 512             for vdu_delta in delta.get("vdu-delta", ()):
 513                 all_monitored_vdus.add(vdu_delta.get("id"))
 514
 515         monitored_vdurs = list(
 516             filter(lambda vdur: vdur["vdu-id-ref"] in all_monitored_vdus, vdurs)
 517         )
 518
 519         if not monitored_vdurs:
 520             log.warning(
 521                 "Scaling criteria is referring to a vnf-monitoring-param that does not "
 522                 "contain a reference to a vdu or vnf metric."
 523             )
 524         return monitored_vdurs
 525
 526     def _get_metric_name(self, vnf_monitoring_param: dict):
 527         if "performance-metric" in vnf_monitoring_param:
 528             return vnf_monitoring_param["performance-metric"]
 529         raise ValueError(
 530             "No metric name found for vnf_monitoring_param %s"
 531             % vnf_monitoring_param["id"]
 532         )