This repository has no description
4.2 kB
99 lines
1# vmalert rules. metrics runs against VictoriaMetrics, logs (LogsQL) against
2# VictoriaLogs; vmalert needs one instance per datasource type.
3{ lib }:
4let
5 inherit (import ./default.nix) alwaysOn;
6
7 alert = name: expr: for: summary: {
8 alert = name;
9 inherit expr for;
10 labels.severity = "warning";
11 annotations.summary = summary;
12 };
13in
14{
15 metrics.groups = [
16 {
17 name = "hosts";
18 rules = [
19 # absent_over_time keeps the host label of its selector, so one term
20 # per machine yields one alert per missing machine.
21 (alert "HostDown" (lib.concatMapStringsSep " or " (
22 host: ''absent_over_time(host_uptime{host="${host}"}[10m])''
23 ) alwaysOn) "0m" "{{ $labels.host }} sent no metrics for 10 minutes")
24 # Network mounts (and their autofs placeholders) are reported by
25 # every machine that mounts them; the agent drops tmpfs & co.
26 # neoprism's /var/download is always above 90%, see DownloadSpace*.
27 (alert "DiskAlmostFull"
28 ''host_filesystem_used_ratio{filesystem!~"cifs|smb3|nfs|nfs4|autofs|fuse.sshfs"} > 0.9 unless on (host, mountpoint) host_filesystem_used_ratio{host="neoprism",mountpoint="/var/download"}''
29 "15m"
30 "{{ $labels.host }}: {{ $labels.mountpoint }} is {{ $value | humanizePercentage }} full"
31 )
32 (alert "DownloadSpaceLow"
33 ''host_filesystem_free_bytes{host="neoprism",mountpoint="/var/download"} < 200e9''
34 "15m"
35 "{{ $labels.host }}: {{ $labels.mountpoint }} has only {{ $value | humanize }}B free"
36 )
37 (
38 alert "DownloadSpaceCritical"
39 ''host_filesystem_free_bytes{host="neoprism",mountpoint="/var/download"} < 100e9''
40 "15m"
41 "{{ $labels.host }}: {{ $labels.mountpoint }} has only {{ $value | humanize }}B free"
42 // {
43 labels.severity = "critical";
44 }
45 )
46 (alert "HighMemory" "1 - host_memory_available_bytes / host_memory_total_bytes > 0.95" "15m"
47 "{{ $labels.host }} uses {{ $value | humanizePercentage }} of its memory"
48 )
49 (alert "MemoryPressure" "rate(host_pressure_memory_some_seconds_total[10m]) > 0.1" "5m"
50 "{{ $labels.host }}: tasks stalled on memory {{ $value | humanizePercentage }} of the time"
51 )
52 (alert "OOMKill" "increase(host_memory_oom_kill_events_total[10m]) > 0" "0m"
53 "{{ $labels.host }}: {{ $value }} processes OOM-killed in the last 10 minutes"
54 )
55 # NVMe drives set critical_warning bit 0x04 (and so fail smart_status)
56 # once they exceed their rated endurance; their wear is covered by
57 # NvmeSpareLow instead. 251 masks out bit 0x04.
58 (alert "SmartFailed"
59 ''smartctl_device_smart_status == 0 unless on (host, device) smartctl_device{protocol="NVMe"}''
60 "5m"
61 "{{ $labels.host }}: {{ $labels.device }} fails its SMART health check"
62 )
63 (alert "NvmeCriticalWarning" "bitmap_and(smartctl_device_critical_warning, 251) != 0" "5m"
64 "{{ $labels.host }}: {{ $labels.device }} reports critical warning {{ $value }}"
65 )
66 (alert "NvmeMediaErrors" "increase(smartctl_device_media_errors[1h]) > 0" "0m"
67 "{{ $labels.host }}: {{ $labels.device }} had {{ $value }} new media errors in the last hour"
68 )
69 (alert "NvmeSpareLow" "smartctl_device_available_spare < smartctl_device_available_spare_threshold"
70 "5m"
71 "{{ $labels.host }}: {{ $labels.device }} has only {{ $value }}% spare left"
72 )
73 ];
74 }
75 {
76 name = "gatus";
77 rules = [
78 (alert "GatusEndpointDown" "gatus_results_endpoint_success == 0" "5m"
79 "{{ $labels.name }} fails as seen from {{ $labels.vantage }}"
80 )
81 ];
82 }
83 ];
84
85 logs.groups = [
86 {
87 name = "journal";
88 type = "vlogs";
89 interval = "1m";
90 rules = [
91 (alert "SystemdUnitFailed"
92 ''_time:15m "Failed with result" | stats by (host, unit) count() as failures''
93 "0m"
94 "{{ $labels.unit }} failed on {{ $labels.host }}"
95 )
96 ];
97 }
98 ];
99}