X-Git-Url: https://osm.etsi.org/gitweb/?a=blobdiff_plain;f=osm_policy_module%2Fcore%2Fagent.py;h=65c741fea676a8dcd8ab15c51b7ec3dcf2dce14a;hb=10be7c984475707be716708dfdd2d91a655158dc;hp=8a1a31ca850d5b025a560fbfc28c6a18378214aa;hpb=023473295ef70215b9603e1fa0de8d71c0f6d0a2;p=osm%2FPOL.git diff --git a/osm_policy_module/core/agent.py b/osm_policy_module/core/agent.py index 8a1a31c..65c741f 100644 --- a/osm_policy_module/core/agent.py +++ b/osm_policy_module/core/agent.py @@ -25,84 +25,81 @@ import asyncio import datetime import json import logging -from json import JSONDecodeError -import yaml -from aiokafka import AIOKafkaConsumer +import peewee from osm_policy_module.common.common_db_client import CommonDbClient from osm_policy_module.common.lcm_client import LcmClient +from osm_policy_module.common.message_bus_client import MessageBusClient from osm_policy_module.common.mon_client import MonClient from osm_policy_module.core import database from osm_policy_module.core.config import Config from osm_policy_module.core.database import ScalingGroup, ScalingAlarm, ScalingPolicy, ScalingCriteria, DatabaseManager +from osm_policy_module.core.exceptions import VdurNotFound +from osm_policy_module.utils.vnfd import VnfdUtils log = logging.getLogger(__name__) -ALLOWED_KAFKA_KEYS = ['instantiated', 'scaled', 'notify_alarm'] +ALLOWED_KAFKA_KEYS = ['instantiated', 'scaled', 'terminated', 'notify_alarm'] class PolicyModuleAgent: - def __init__(self, loop=None): - cfg = Config.instance() + def __init__(self, config: Config, loop=None): + self.conf = config if not loop: loop = asyncio.get_event_loop() self.loop = loop - self.db_client = CommonDbClient() - self.mon_client = MonClient(loop=self.loop) - self.lcm_client = LcmClient(loop=self.loop) - self.kafka_server = '{}:{}'.format(cfg.OSMPOL_MESSAGE_HOST, - cfg.OSMPOL_MESSAGE_PORT) - self.database_manager = DatabaseManager() + self.db_client = CommonDbClient(config) + self.mon_client = MonClient(config, loop=self.loop) + self.lcm_client = LcmClient(config, loop=self.loop) + self.database_manager = DatabaseManager(config) + self.msg_bus = MessageBusClient(config) def run(self): self.loop.run_until_complete(self.start()) async def start(self): - consumer = AIOKafkaConsumer( + topics = [ "ns", - "alarm_response", - loop=self.loop, - bootstrap_servers=self.kafka_server, - group_id="pol-consumer", - key_deserializer=bytes.decode, - value_deserializer=bytes.decode, - ) - await consumer.start() - try: - async for msg in consumer: - await self._process_msg(msg.topic, msg.key, msg.value) - finally: - await consumer.stop() + "alarm_response" + ] + await self.msg_bus.aioread(topics, self._process_msg) + log.critical("Exiting...") async def _process_msg(self, topic, key, msg): log.debug("_process_msg topic=%s key=%s msg=%s", topic, key, msg) + log.info("Message arrived: %s", msg) try: if key in ALLOWED_KAFKA_KEYS: - try: - content = json.loads(msg) - except JSONDecodeError: - content = yaml.safe_load(msg) - if key == 'instantiated' or key == 'scaled': - await self._handle_instantiated_or_scaled(content) + if key == 'instantiated': + await self._handle_instantiated(msg) + + if key == 'scaled': + await self._handle_scaled(msg) + + if key == 'terminated': + await self._handle_terminated(msg) if key == 'notify_alarm': - await self._handle_alarm_notification(content) + await self._handle_alarm_notification(msg) else: log.debug("Key %s is not in ALLOWED_KAFKA_KEYS", key) + except peewee.PeeweeException: + log.exception("Database error consuming message: ") + raise except Exception: log.exception("Error consuming message: ") async def _handle_alarm_notification(self, content): log.debug("_handle_alarm_notification: %s", content) - alarm_id = content['notify_details']['alarm_uuid'] + alarm_uuid = content['notify_details']['alarm_uuid'] metric_name = content['notify_details']['metric_name'] operation = content['notify_details']['operation'] threshold = content['notify_details']['threshold_value'] vdu_name = content['notify_details']['vdu_name'] vnf_member_index = content['notify_details']['vnf_member_index'] - ns_id = content['notify_details']['ns_id'] + nsr_id = content['notify_details']['ns_id'] log.info( "Received alarm notification for alarm %s, \ metric %s, \ @@ -111,9 +108,9 @@ class PolicyModuleAgent: vdu_name %s, \ vnf_member_index %s, \ ns_id %s ", - alarm_id, metric_name, operation, threshold, vdu_name, vnf_member_index, ns_id) + alarm_uuid, metric_name, operation, threshold, vdu_name, vnf_member_index, nsr_id) try: - alarm = self.database_manager.get_alarm(alarm_id) + alarm = self.database_manager.get_alarm(alarm_uuid) delta = datetime.datetime.now() - alarm.scaling_criteria.scaling_policy.last_scale log.debug("last_scale: %s", alarm.scaling_criteria.scaling_policy.last_scale) log.debug("now: %s", datetime.datetime.now()) @@ -121,33 +118,60 @@ class PolicyModuleAgent: if delta.total_seconds() < alarm.scaling_criteria.scaling_policy.cooldown_time: log.info("Time between last scale and now is less than cooldown time. Skipping.") return - log.info("Sending scaling action message for ns: %s", alarm_id) - await self.lcm_client.scale(ns_id, + log.info("Sending scaling action message for ns: %s", nsr_id) + await self.lcm_client.scale(nsr_id, alarm.scaling_criteria.scaling_policy.scaling_group.name, alarm.vnf_member_index, alarm.action) alarm.scaling_criteria.scaling_policy.last_scale = datetime.datetime.now() alarm.scaling_criteria.scaling_policy.save() except ScalingAlarm.DoesNotExist: - log.info("There is no action configured for alarm %s.", alarm_id) + log.info("There is no action configured for alarm %s.", alarm_uuid) + + async def _handle_instantiated(self, content): + log.debug("_handle_instantiated: %s", content) + nslcmop_id = content['nslcmop_id'] + nslcmop = self.db_client.get_nslcmop(nslcmop_id) + if nslcmop['operationState'] == 'COMPLETED' or nslcmop['operationState'] == 'PARTIALLY_COMPLETED': + nsr_id = nslcmop['nsInstanceId'] + log.info("Configuring scaling groups for network service with nsr_id: %s", nsr_id) + await self._configure_scaling_groups(nsr_id) + else: + log.info( + "Network service is not in COMPLETED or PARTIALLY_COMPLETED state. " + "Current state is %s. Skipping...", + nslcmop['operationState']) - async def _handle_instantiated_or_scaled(self, content): - log.debug("_handle_instantiated_or_scaled: %s", content) + async def _handle_scaled(self, content): + log.debug("_handle_scaled: %s", content) nslcmop_id = content['nslcmop_id'] nslcmop = self.db_client.get_nslcmop(nslcmop_id) if nslcmop['operationState'] == 'COMPLETED' or nslcmop['operationState'] == 'PARTIALLY_COMPLETED': nsr_id = nslcmop['nsInstanceId'] log.info("Configuring scaling groups for network service with nsr_id: %s", nsr_id) await self._configure_scaling_groups(nsr_id) + log.info("Checking for orphaned alarms to be deleted for network service with nsr_id: %s", nsr_id) + await self._delete_orphaned_alarms(nsr_id) else: log.info( "Network service is not in COMPLETED or PARTIALLY_COMPLETED state. " "Current state is %s. Skipping...", nslcmop['operationState']) + async def _handle_terminated(self, content): + log.debug("_handle_deleted: %s", content) + nsr_id = content['nsr_id'] + if content['operationState'] == 'COMPLETED' or content['operationState'] == 'PARTIALLY_COMPLETED': + log.info("Deleting scaling groups and alarms for network service with nsr_id: %s", nsr_id) + await self._delete_scaling_groups(nsr_id) + else: + log.info( + "Network service is not in COMPLETED or PARTIALLY_COMPLETED state. " + "Current state is %s. Skipping...", + content['operationState']) + async def _configure_scaling_groups(self, nsr_id: str): log.debug("_configure_scaling_groups: %s", nsr_id) - # TODO: Add support for non-nfvi metrics alarms_created = [] with database.db.atomic() as tx: try: @@ -247,8 +271,13 @@ class PolicyModuleAgent: ) ) elif 'vnf-metric' in vnf_monitoring_param: - log.warning("vnf-metric is not currently supported.") - continue + vdu = VnfdUtils.get_mgmt_vdu(vnfd) + vdurs = list( + filter( + lambda vdur: vdur['vdu-id-ref'] == vdu['id'], + vnfr['vdur'] + ) + ) else: log.warning( "Scaling criteria is referring to a vnf-monitoring-param that does not " @@ -280,13 +309,14 @@ class PolicyModuleAgent: operation=scaling_criteria['scale-in-relational-operation'], statistic=vnf_monitoring_param['aggregation-type'] ) - ScalingAlarm.create( - alarm_id=alarm_uuid, + alarm = ScalingAlarm.create( + alarm_uuid=alarm_uuid, action='scale_in', vnf_member_index=int(vnfr['member-vnf-index-ref']), vdu_name=vdur['name'], scaling_criteria=scaling_criteria_record ) + alarms_created.append(alarm) alarm_uuid = await self.mon_client.create_alarm( metric_name=vnf_monitoring_param['id'], ns_id=nsr_id, @@ -296,13 +326,14 @@ class PolicyModuleAgent: operation=scaling_criteria['scale-out-relational-operation'], statistic=vnf_monitoring_param['aggregation-type'] ) - ScalingAlarm.create( - alarm_id=alarm_uuid, + alarm = ScalingAlarm.create( + alarm_uuid=alarm_uuid, action='scale_out', vnf_member_index=int(vnfr['member-vnf-index-ref']), vdu_name=vdur['name'], scaling_criteria=scaling_criteria_record ) + alarms_created.append(alarm) except Exception as e: log.exception("Error configuring scaling groups:") @@ -310,5 +341,59 @@ class PolicyModuleAgent: if len(alarms_created) > 0: log.info("Cleaning alarm resources in MON") for alarm in alarms_created: - await self.mon_client.delete_alarm(*alarm) + await self.mon_client.delete_alarm(alarm.scaling_criteria.scaling_policy.scaling_group.nsr_id, + alarm.vnf_member_index, + alarm.vdu_name, + alarm.alarm_uuid) + raise e + + async def _delete_scaling_groups(self, nsr_id: str): + with database.db.atomic() as tx: + try: + for scaling_group in ScalingGroup.select().where(ScalingGroup.nsr_id == nsr_id): + for scaling_policy in scaling_group.scaling_policies: + for scaling_criteria in scaling_policy.scaling_criterias: + for alarm in scaling_criteria.scaling_alarms: + try: + await self.mon_client.delete_alarm( + alarm.scaling_criteria.scaling_policy.scaling_group.nsr_id, + alarm.vnf_member_index, + alarm.vdu_name, + alarm.alarm_uuid) + except ValueError: + log.exception("Error deleting alarm in MON %s", alarm.alarm_uuid) + alarm.delete_instance() + scaling_criteria.delete_instance() + scaling_policy.delete_instance() + scaling_group.delete_instance() + + except Exception as e: + log.exception("Error deleting scaling groups and alarms:") + tx.rollback() + raise e + + async def _delete_orphaned_alarms(self, nsr_id): + with database.db.atomic() as tx: + try: + for scaling_group in ScalingGroup.select().where(ScalingGroup.nsr_id == nsr_id): + for scaling_policy in scaling_group.scaling_policies: + for scaling_criteria in scaling_policy.scaling_criterias: + for alarm in scaling_criteria.scaling_alarms: + try: + self.db_client.get_vdur(nsr_id, alarm.vnf_member_index, alarm.vdu_name) + except VdurNotFound: + log.info("Deleting orphaned alarm %s", alarm.alarm_uuid) + try: + await self.mon_client.delete_alarm( + alarm.scaling_criteria.scaling_policy.scaling_group.nsr_id, + alarm.vnf_member_index, + alarm.vdu_name, + alarm.alarm_uuid) + except ValueError: + log.exception("Error deleting alarm in MON %s", alarm.alarm_uuid) + alarm.delete_instance() + + except Exception as e: + log.exception("Error deleting orphaned alarms:") + tx.rollback() raise e