mirror of
https://github.com/monitoring-mixins/website.git
synced 2024-12-14 11:37:31 +00:00
114 lines
3.8 KiB
YAML
114 lines
3.8 KiB
YAML
groups:
|
|
- name: MongodbAlerts
|
|
rules:
|
|
- alert: MongodbDown
|
|
annotations:
|
|
description: |-
|
|
MongoDB instance is down
|
|
VALUE = {{ $value }}
|
|
LABELS = {{ $labels }}
|
|
summary: MongoDB Instance is Down.
|
|
expr: mongodb_up == 0
|
|
for: 5m
|
|
labels:
|
|
severity: critical
|
|
- alert: MongodbReplicaMemberUnhealthy
|
|
annotations:
|
|
description: Mongodb replica member unhealthy (instance {{ $labels.instance
|
|
}})
|
|
summary: MongoDB replica member unhealthy.
|
|
expr: mongodb_mongod_replset_member_health == 0
|
|
labels:
|
|
severity: critical
|
|
- alert: MongodbReplicationLag
|
|
annotations:
|
|
description: |-
|
|
Mongodb replication lag is more than 10s
|
|
VALUE = {{ $value }}
|
|
LABELS = {{ $labels }}
|
|
summary: MongoDB replication lag is exceeding the threshold.
|
|
expr: mongodb_mongod_replset_member_replication_lag{state="SECONDARY"} > 60
|
|
for: 5m
|
|
labels:
|
|
severity: critical
|
|
- alert: MongodbReplicationHeadroom
|
|
annotations:
|
|
description: |-
|
|
MongoDB replication headroom is <= 0
|
|
VALUE = {{ $value }}
|
|
LABELS = {{ $labels }}
|
|
summary: MongoDB replication headroom is exceeding the threshold.
|
|
expr: (avg(mongodb_mongod_replset_oplog_tail_timestamp - mongodb_mongod_replset_oplog_head_timestamp)
|
|
- (avg(mongodb_mongod_replset_member_optime_date{state="PRIMARY"}) - avg(mongodb_mongod_replset_member_optime_date{state="SECONDARY"})))
|
|
<= 0
|
|
for: 5m
|
|
labels:
|
|
severity: critical
|
|
- alert: MongodbNumberCursorsOpen
|
|
annotations:
|
|
description: |-
|
|
Too many cursors opened by MongoDB for clients (> 10k)
|
|
VALUE = {{ $value }}
|
|
LABELS = {{ $labels }}
|
|
summary: MongoDB number cursors open too high.
|
|
expr: mongodb_mongod_metrics_cursor_open{state="total"} > 10 * 1000
|
|
for: 2m
|
|
labels:
|
|
severity: warning
|
|
- alert: MongodbCursorsTimeouts
|
|
annotations:
|
|
description: |-
|
|
Too many cursors are timing out
|
|
VALUE = {{ $value }}
|
|
LABELS = {{ $labels }}
|
|
summary: MongoDB cursors timeouts is exceeding the threshold.
|
|
expr: increase(mongodb_mongod_metrics_cursor_timed_out_total[1m]) > 100
|
|
for: 2m
|
|
labels:
|
|
severity: warning
|
|
- alert: MongodbTooManyConnections
|
|
annotations:
|
|
description: |-
|
|
Too many connections (> 80%)
|
|
VALUE = {{ $value }}
|
|
LABELS = {{ $labels }}
|
|
summary: MongoDB too many connections.
|
|
expr: avg by(instance) (rate(mongodb_connections{state="current"}[1m])) / avg
|
|
by(instance) (sum (mongodb_connections) by (instance)) * 100 > 80
|
|
for: 2m
|
|
labels:
|
|
severity: warning
|
|
- alert: MongodbVirtualMemoryUsage
|
|
annotations:
|
|
description: |-
|
|
High memory usage
|
|
VALUE = {{ $value }}
|
|
LABELS = {{ $labels }}
|
|
summary: MongoDB virtual memory usage is too high.
|
|
expr: (sum(mongodb_memory{type="virtual"}) BY (instance) / sum(mongodb_memory{type="mapped"})
|
|
BY (instance)) > 3
|
|
for: 2m
|
|
labels:
|
|
severity: warning
|
|
- alert: MongodbReadRequestsQueueingUp
|
|
annotations:
|
|
description: |-
|
|
MongoDB requests are queuing up
|
|
VALUE = {{ $value }}
|
|
LABELS = {{ $labels }}
|
|
summary: MongoDB read requests queuing up.
|
|
expr: delta(mongodb_mongod_global_lock_current_queue{type="reader"}[1m]) > 0
|
|
for: 5m
|
|
labels:
|
|
severity: warning
|
|
- alert: MongodbWriteRequestsQueueingUp
|
|
annotations:
|
|
description: |-
|
|
MongoDB write requests are queueing up
|
|
VALUE = {{ $value }}
|
|
LABELS = {{ $labels }}
|
|
summary: MongoDB write requests queuing up.
|
|
expr: delta(mongodb_mongod_global_lock_current_queue{type="writer"}[1m]) > 0
|
|
for: 5m
|
|
labels:
|
|
severity: warning
|