feat(health): add charts.d/nut alarms (#12285)
Ilya Mashchenko committed
Mar 2, 2022 at 12:50 UTC
57ebc01370d4b67a654db6b1a9b2ff2365956d57
2 files changed
+48
health/Makefile.am
+1
@@ -65,6 +65,7 @@ dist_healthconfig_DATA = \
65
health.d/mysql.conf \
66
health.d/net.conf \
67
health.d/netfilter.conf \
68
+ health.d/nut.conf \
69
health.d/pihole.conf \
70
health.d/portcheck.conf \
71
health.d/processes.conf \
health/health.d/nut.conf
new
+47
@@ -0,0 +1,47 @@
1
+# you can disable an alarm notification by setting the 'to' line to: silent
2
+
3
+ template: nut_10min_ups_load
4
+ on: nut.load
5
+ class: Utilization
6
+ type: Power Supply
7
+component: UPS
8
+ os: *
9
+ hosts: *
10
+ lookup: average -10m unaligned of load
11
+ units: %
12
+ every: 1m
13
+ warn: $this > (($status >= $WARNING) ? (70) : (80))
14
+ crit: $this > (($status == $CRITICAL) ? (85) : (95))
15
+ delay: down 10m multiplier 1.5 max 1h
16
+ info: average UPS load over the last 10 minutes
17
+ to: sitemgr
18
+
19
+ template: nut_ups_charge
20
+ on: nut.charge
21
+ class: Errors
22
+ type: Power Supply
23
+component: UPS
24
+ os: *
25
+ hosts: *
26
+ lookup: average -60s unaligned of battery_charge
27
+ units: %
28
+ every: 60s
29
+ warn: $this < 100
30
+ crit: $this < (($status == $CRITICAL) ? (60) : (50))
31
+ delay: down 10m multiplier 1.5 max 1h
32
+ info: average UPS charge over the last minute
33
+ to: sitemgr
34
+
35
+ template: nut_last_collected_secs
36
+ on: nut.load
37
+ class: Latency
38
+ type: Power Supply
39
+component: UPS device
40
+ calc: $now - $last_collected_t
41
+ every: 10s
42
+ units: seconds ago
43
+ warn: $this > (($status >= $WARNING) ? ($update_every) : ( 5 * $update_every))
44
+ crit: $this > (($status == $CRITICAL) ? ($update_every) : (60 * $update_every))
45
+ delay: down 5m multiplier 1.5 max 1h
46
+ info: number of seconds since the last successful data collection
47
+ to: sitemgr