osm_mon/evaluator/service.py

   1 # -*- coding: utf-8 -*-
   2
   3 # Copyright 2018 Whitestack, LLC
   4 # *************************************************************
   5
   6 # This file is part of OSM Monitoring module
   7 # All Rights Reserved to Whitestack, LLC
   8
   9 # Licensed under the Apache License, Version 2.0 (the "License"); you may
  10 # not use this file except in compliance with the License. You may obtain
  11 # a copy of the License at
  12
  13 #         http://www.apache.org/licenses/LICENSE-2.0
  14
  15 # Unless required by applicable law or agreed to in writing, software
  16 # distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
  17 # WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
  18 # License for the specific language governing permissions and limitations
  19 # under the License.
  20 # For those usages not covered by the Apache License, Version 2.0 please
  21 # contact: bdiaz@whitestack.com or glavado@whitestack.com
  22 ##
  23 import logging
  24 import multiprocessing
  25 from enum import Enum
  26 from typing import Tuple, List
  27
  28 from osm_mon.core.common_db import CommonDbClient
  29 from osm_mon.core.config import Config
  30 from osm_mon.core.models import Alarm
  31 from osm_mon.evaluator.backends.prometheus import PrometheusBackend
  32
  33 log = logging.getLogger(__name__)
  34
  35 BACKENDS = {"prometheus": PrometheusBackend}
  36
  37
  38 class AlarmStatus(Enum):
  39     ALARM = "alarm"
  40     OK = "ok"
  41     INSUFFICIENT = "insufficient-data"
  42     DISABLED = "disabled"
  43
  44
  45 class EvaluatorService:
  46     def __init__(self, config: Config):
  47         self.conf = config
  48         self.common_db = CommonDbClient(self.conf)
  49         self.queue = multiprocessing.Queue()
  50
  51     def _get_metric_value(self, metric_name: str, tags: dict):
  52         return BACKENDS[self.conf.get("evaluator", "backend")](
  53             self.conf
  54         ).get_metric_value(metric_name, tags)
  55
  56     def _evaluate_metric(self, alarm: Alarm):
  57         """Method to evaluate a metric value comparing it against an alarm threshold.
  58
  59         Args:
  60             alarm (Alarm): the alarm with the threshold to compare the metric against
  61         """
  62
  63         log.debug("_evaluate_metric")
  64         metric_value = self._get_metric_value(alarm.metric, alarm.tags)
  65         if alarm.alarm_status.upper() != AlarmStatus.DISABLED.value.upper():
  66             if metric_value is None:
  67                 log.warning("No metric result for alarm %s", alarm.uuid)
  68                 self.queue.put((alarm, AlarmStatus.INSUFFICIENT))
  69             else:
  70                 if (
  71                     (alarm.operation.upper() == "GT" and metric_value > alarm.threshold)
  72                     or (
  73                         alarm.operation.upper() == "LT"
  74                         and metric_value < alarm.threshold
  75                     )
  76                     or (
  77                         alarm.operation.upper() == "GE"
  78                         and metric_value >= alarm.threshold
  79                     )
  80                     or (
  81                         alarm.operation.upper() == "LE"
  82                         and metric_value <= alarm.threshold
  83                     )
  84                     or (
  85                         alarm.operation.upper() == "EQ"
  86                         and metric_value == alarm.threshold
  87                     )
  88                     or (
  89                         alarm.operation.upper() == "NE"
  90                         and metric_value != alarm.threshold
  91                     )
  92                 ):
  93                     self.queue.put((alarm, AlarmStatus.ALARM))
  94                 elif alarm.operation.upper() in ("GT", "LT", "GE", "LE", "EQ", "NE"):
  95                     self.queue.put((alarm, AlarmStatus.OK))
  96
  97     def update_alarm_status(self, alarm_state, uuid):
  98         alarm_data = self.common_db.get_alarm_by_uuid(uuid)
  99         if alarm_data.get("alarm_status").upper() != AlarmStatus.DISABLED.value.upper():
 100             self.common_db.update_alarm_status(alarm_state, uuid)
 101         return
 102
 103     def evaluate_alarms(self) -> List[Tuple[Alarm, AlarmStatus]]:
 104         log.debug("evaluate_alarms")
 105         processes = []
 106         for alarm in self.common_db.get_alarms():
 107             p = multiprocessing.Process(target=self._evaluate_metric, args=(alarm,))
 108             processes.append(p)
 109             p.start()
 110
 111         for process in processes:
 112             process.join(timeout=10)
 113         alarms_tuples = []
 114         while not self.queue.empty():
 115             alarms_tuples.append(self.queue.get())
 116         return alarms_tuples