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