e97586ce7cd146d0a0d9d1ca515eaf2e48a87ecb
1 # -*- coding: utf-8 -*-
3 # Copyright 2018 Whitestack, LLC
4 # *************************************************************
6 # This file is part of OSM Monitoring module
7 # All Rights Reserved to Whitestack, LLC
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
13 # http://www.apache.org/licenses/LICENSE-2.0
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
20 # For those usages not covered by the Apache License, Version 2.0 please
21 # contact: bdiaz@whitestack.com or glavado@whitestack.com
24 import multiprocessing
26 from typing
import Tuple
, List
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
33 log
= logging
.getLogger(__name__
)
35 BACKENDS
= {"prometheus": PrometheusBackend
}
38 class AlarmStatus(Enum
):
41 INSUFFICIENT
= "insufficient-data"
45 class EvaluatorService
:
46 def __init__(self
, config
: Config
):
48 self
.common_db
= CommonDbClient(self
.conf
)
49 self
.queue
= multiprocessing
.Queue()
51 def _get_metric_data(self
, metric_name
: str, tags
: dict):
52 return BACKENDS
[self
.conf
.get("evaluator", "backend")](
54 ).get_metric_data(metric_name
, tags
)
56 def _evaluate_metric(self
, alarm
: Alarm
):
57 """Method to evaluate a metric value comparing it against an alarm threshold.
60 alarm (Alarm): the alarm with the threshold to compare the metric against
63 log
.debug("_evaluate_metric")
64 metric_data
= self
._get
_metric
_data
(alarm
.metric
, alarm
.tags
)
65 if metric_data
is None:
66 log
.warning("No metric result for alarm %s", alarm
.uuid
)
67 self
.queue
.put((alarm
, AlarmStatus
.INSUFFICIENT
))
69 for metric
in metric_data
:
70 metric_value
= metric
["value"]
71 metric_labels
= metric
["labels"]
72 alarm
.extra_labels
.update(metric_labels
)
73 if alarm
.alarm_status
.upper() != AlarmStatus
.DISABLED
.value
.upper():
76 alarm
.operation
.upper() == "GT"
77 and metric_value
> alarm
.threshold
80 alarm
.operation
.upper() == "LT"
81 and metric_value
< alarm
.threshold
84 alarm
.operation
.upper() == "GE"
85 and metric_value
>= alarm
.threshold
88 alarm
.operation
.upper() == "LE"
89 and metric_value
<= alarm
.threshold
92 alarm
.operation
.upper() == "EQ"
93 and metric_value
== alarm
.threshold
96 alarm
.operation
.upper() == "NE"
97 and metric_value
!= alarm
.threshold
100 self
.queue
.put((alarm
, AlarmStatus
.ALARM
))
101 elif alarm
.operation
.upper() in (
109 self
.queue
.put((alarm
, AlarmStatus
.OK
))
111 def update_alarm_status(self
, alarm_state
, uuid
):
112 alarm_data
= self
.common_db
.get_alarm_by_uuid(uuid
)
113 if alarm_data
.get("alarm_status").upper() != AlarmStatus
.DISABLED
.value
.upper():
114 self
.common_db
.update_alarm_status(alarm_state
, uuid
)
117 def update_alarm_extra_labels(self
, alarm_labels
, uuid
):
118 self
.common_db
.update_alarm_extra_labels(alarm_labels
, uuid
)
121 def evaluate_alarms(self
) -> List
[Tuple
[Alarm
, AlarmStatus
]]:
122 log
.debug("evaluate_alarms")
124 for alarm
in self
.common_db
.get_alarms():
125 p
= multiprocessing
.Process(target
=self
._evaluate
_metric
, args
=(alarm
,))
129 for process
in processes
:
130 process
.join(timeout
=10)
132 while not self
.queue
.empty():
133 alarms_tuples
.append(self
.queue
.get())