X-Git-Url: https://osm.etsi.org/gitweb/?a=blobdiff_plain;f=osm_policy_module%2Fcore%2Fagent.py;h=4a2052760db192a2bc17fb7235359af8d68cc2f6;hb=055c4ee00f83647e7e807aa3d44c3384d9d79aa2;hp=52412a6d8ae54a9956adaaeed6cfdbf9fc142157;hpb=be75541d2f2306f8992f47bca8179210f858c593;p=osm%2FPOL.git diff --git a/osm_policy_module/core/agent.py b/osm_policy_module/core/agent.py index 52412a6..4a20527 100644 --- a/osm_policy_module/core/agent.py +++ b/osm_policy_module/core/agent.py @@ -21,159 +21,202 @@ # For those usages not covered by the Apache License, Version 2.0 please # contact: bdiaz@whitestack.com or glavado@whitestack.com ## -import json +import asyncio import logging -import threading -from json import JSONDecodeError +from pathlib import Path +import os -import yaml -from kafka import KafkaConsumer +import peewee -from osm_policy_module.common.db_client import DbClient -from osm_policy_module.common.lcm_client import LcmClient -from osm_policy_module.common.mon_client import MonClient -from osm_policy_module.core import database +from osm_policy_module.alarming.service import AlarmingService +from osm_policy_module.autoscaling.service import AutoscalingService +from osm_policy_module.healing.service import HealingService +from osm_policy_module.common.common_db_client import CommonDbClient +from osm_policy_module.common.message_bus_client import MessageBusClient from osm_policy_module.core.config import Config -from osm_policy_module.core.database import ScalingRecord, ScalingAlarm log = logging.getLogger(__name__) +ALLOWED_KAFKA_KEYS = [ + "instantiated", + "scaled", + "terminated", + "notify_alarm", + "policy_updated", + "vnf_terminated", +] + class PolicyModuleAgent: - def __init__(self): - cfg = Config.instance() - self.db_client = DbClient() - self.mon_client = MonClient() - self.kafka_server = '{}:{}'.format(cfg.OSMPOL_MESSAGE_HOST, - cfg.OSMPOL_MESSAGE_PORT) + def __init__(self, config: Config): + self.conf = config + self.msg_bus = MessageBusClient(config) + self.db_client = CommonDbClient(config) + self.autoscaling_service = AutoscalingService(config) + self.alarming_service = AlarmingService(config) + self.healing_service = HealingService(config) def run(self): - consumer = KafkaConsumer(bootstrap_servers=self.kafka_server, - key_deserializer=bytes.decode, - value_deserializer=bytes.decode, - group_id='pol-consumer') - consumer.subscribe(["ns", "alarm_response"]) + asyncio.run(self.start()) + + async def start(self): + Path("/tmp/osm_pol_agent_health_flag").touch() + topics = ["ns", "alarm_response"] + await self.msg_bus.aioread(topics, self._process_msg) + log.critical("Exiting...") + if os.path.exists("/tmp/osm_pol_agent_health_flag"): + os.remove("/tmp/osm_pol_agent_health_flag") + + async def _process_msg(self, topic, key, msg): + Path("/tmp/osm_pol_agent_health_flag").touch() + log.debug("_process_msg topic=%s key=%s msg=%s", topic, key, msg) + try: + if key in ALLOWED_KAFKA_KEYS: + if key == "instantiated": + await self._handle_instantiated(msg) - for message in consumer: - t = threading.Thread(target=self._process_msg, args=(message.topic, message.key, message.value,)) - t.start() + if key == "scaled": + await self._handle_scaled(msg) - def _process_msg(self, topic, key, msg): - try: - # Check for ns instantiation - if key == 'instantiated': - try: - content = json.loads(msg) - except JSONDecodeError: - content = yaml.safe_load(msg) - log.info("Message arrived with topic: %s, key: %s, msg: %s", topic, key, 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) - 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']) - - if key == 'notify_alarm': - try: - content = json.loads(msg) - except JSONDecodeError: - content = yaml.safe_load(msg) - log.info("Message arrived with topic: %s, key: %s, msg: %s", topic, key, content) - alarm_id = 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'] - log.info( - "Received alarm notification for alarm %s, \ - metric %s, \ - operation %s, \ - threshold %s, \ - vdu_name %s, \ - vnf_member_index %s, \ - ns_id %s ", - alarm_id, metric_name, operation, threshold, vdu_name, vnf_member_index, ns_id) - try: - alarm = ScalingAlarm.select().where(ScalingAlarm.alarm_id == alarm_id).get() - lcm_client = LcmClient() - log.info("Sending scaling action message for ns: %s", alarm_id) - lcm_client.scale(alarm.scaling_record.nsr_id, alarm.scaling_record.name, alarm.vnf_member_index, - alarm.action) - except ScalingAlarm.DoesNotExist: - log.info("There is no action configured for alarm %s.", alarm_id) + if key == "terminated": + await self._handle_terminated(msg) + + if key == "notify_alarm": + await self._handle_alarm_notification(msg) + + if key == "policy_updated": + await self._handle_policy_update(msg) + + if key == "vnf_terminated": + await self._handle_vnf_terminated(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: ") - def _configure_scaling_groups(self, nsr_id: str): - # TODO(diazb): Check for alarm creation on exception and clean resources if needed. - with database.db.atomic(): - vnfrs = self.db_client.get_vnfrs(nsr_id) - log.info("Checking %s vnfrs...", len(vnfrs)) - for vnfr in vnfrs: - vnfd = self.db_client.get_vnfd(vnfr['vnfd-id']) - log.info("Looking for vnfd %s", vnfr['vnfd-id']) - scaling_groups = vnfd['scaling-group-descriptor'] - vnf_monitoring_params = vnfd['monitoring-param'] - for scaling_group in scaling_groups: - log.info("Creating scaling record in DB...") - scaling_record = ScalingRecord.create( - nsr_id=nsr_id, - name=scaling_group['name'], - content=json.dumps(scaling_group) - ) - log.info("Created scaling record in DB : nsr_id=%s, name=%s, content=%s", - scaling_record.nsr_id, - scaling_record.name, - scaling_record.content) - for scaling_policy in scaling_group['scaling-policy']: - for vdur in vnfd['vdu']: - vdu_monitoring_params = vdur['monitoring-param'] - for scaling_criteria in scaling_policy['scaling-criteria']: - vnf_monitoring_param = next( - filter(lambda param: param['id'] == scaling_criteria['vnf-monitoring-param-ref'], - vnf_monitoring_params)) - # TODO: Add support for non-nfvi metrics - vdu_monitoring_param = next( - filter( - lambda param: param['id'] == vnf_monitoring_param['vdu-monitoring-param-ref'], - vdu_monitoring_params)) - alarm_uuid = self.mon_client.create_alarm( - metric_name=vdu_monitoring_param['nfvi-metric'], - ns_id=nsr_id, - vdu_name=vdur['name'], - vnf_member_index=vnfr['member-vnf-index-ref'], - threshold=scaling_criteria['scale-in-threshold'], - operation=scaling_criteria['scale-in-relational-operation'], - statistic=vnf_monitoring_param['aggregation-type'] - ) - ScalingAlarm.create( - alarm_id=alarm_uuid, - action='scale_in', - vnf_member_index=int(vnfr['member-vnf-index-ref']), - vdu_name=vdur['name'], - scaling_record=scaling_record - ) - alarm_uuid = self.mon_client.create_alarm( - metric_name=vdu_monitoring_param['nfvi-metric'], - ns_id=nsr_id, - vdu_name=vdur['name'], - vnf_member_index=vnfr['member-vnf-index-ref'], - threshold=scaling_criteria['scale-out-threshold'], - operation=scaling_criteria['scale-out-relational-operation'], - statistic=vnf_monitoring_param['aggregation-type'] - ) - ScalingAlarm.create( - alarm_id=alarm_uuid, - action='scale_out', - vnf_member_index=int(vnfr['member-vnf-index-ref']), - vdu_name=vdur['name'], - scaling_record=scaling_record - ) + async def _handle_alarm_notification(self, content): + log.debug("_handle_alarm_notification: %s", content) + alarm_uuid = content["notify_details"]["alarm_uuid"] + status = content["notify_details"]["status"] + await self.autoscaling_service.handle_alarm(alarm_uuid, status) + await self.alarming_service.handle_alarm(alarm_uuid, status, content) + await self.healing_service.handle_alarm(alarm_uuid, status) + + 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 nsr_id: %s", nsr_id) + await self.autoscaling_service.configure_scaling_groups(nsr_id) + await self.alarming_service.configure_vnf_alarms(nsr_id) + await self.healing_service.configure_healing_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_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 scaled service with nsr_id: %s", nsr_id) + await self.autoscaling_service.configure_scaling_groups(nsr_id) + await self.autoscaling_service.delete_orphaned_alarms(nsr_id) + await self.alarming_service.configure_vnf_alarms(nsr_id) + await self.healing_service.configure_healing_alarms(nsr_id) + await self.healing_service.delete_orphaned_healing_alarms(nsr_id) + else: + log.debug( + "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 autoscaling_service with nsr_id: %s", + nsr_id, + ) + await self.autoscaling_service.delete_scaling_groups(nsr_id) + await self.alarming_service.delete_vnf_alarms(nsr_id) + await self.healing_service.delete_healing_alarms(nsr_id) + else: + log.info( + "Network service is not in COMPLETED or PARTIALLY_COMPLETED state. " + "Current state is %s. Skipping...", + content["operationState"], + ) + + async def _handle_policy_update(self, content): + log.info("_handle_policy_update: %s", content) + nsr_id = content["nsr_id"] + vnf_member_index = content["vnf_member_index"] + if ( + content["operationState"] == "COMPLETED" + or content["operationState"] == "PARTIALLY_COMPLETED" + ): + log.info( + "Updating policies of VNF with nsr_id: %s and vnf-member-index: %s" + % (nsr_id, vnf_member_index) + ) + await self.autoscaling_service.delete_scaling_groups( + nsr_id, vnf_member_index + ) + await self.alarming_service.delete_vnf_alarms(nsr_id, vnf_member_index) + await self.healing_service.delete_healing_alarms(nsr_id, vnf_member_index) + await self.autoscaling_service.configure_scaling_groups( + nsr_id, vnf_member_index + ) + await self.alarming_service.configure_vnf_alarms(nsr_id, vnf_member_index) + await self.healing_service.configure_healing_alarms( + nsr_id, vnf_member_index + ) + else: + log.info( + "Network service is not in COMPLETED or PARTIALLY_COMPLETED state. " + "Current state is %s. Skipping...", + content["operationState"], + ) + + async def _handle_vnf_terminated(self, content): + nsr_id = content["nsr_id"] + vnf_member_index = content["vnf_member_index"] + if ( + content["operationState"] == "COMPLETED" + or content["operationState"] == "PARTIALLY_COMPLETED" + ): + log.info( + "Deleting policies of VNF with nsr_id: %s and vnf-member-index: %s" + % (nsr_id, vnf_member_index) + ) + await self.autoscaling_service.delete_scaling_groups( + nsr_id, vnf_member_index + ) + await self.alarming_service.delete_vnf_alarms(nsr_id, vnf_member_index) + else: + log.info( + "Network service is not in COMPLETED or PARTIALLY_COMPLETED state. " + "Current state is %s. Skipping...", + content["operationState"], + )