newedge/infra
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

mob next [ci-skip] [ci skip] [skip ci]

Browse Source 
lastFile:modules/clan/prometheus/default.nix
This commit is contained in:
kurogeek
2026-06-19 15:41:01 +07:00
parent 6d2e73267f
commit 50622270ec
2 changed files with 15 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files
inventories/default.nix
-17
View File
@@ -360,23 +360,6 @@
};
services.prometheus = {
# rules = [
# (builtins.toJSON {
# groups = [
# {
# name = "default";
# rules = [
# {
# alert = "NodeDown";
# expr = "up == 0";
# for = "1m";
# annotations.summary = "Node {{ $labels.job }} has been down for more than 1 minutes. {{ $labels.instance }}";
# }
# ];
# }
# ];
# })
# ];
alertmanager = {
enable = true;
modules/clan/prometheus/default.nix
+15 -2
View File
@@ -103,11 +103,24 @@
expr = "smartctl_device_error_log_count > 0";
for = "5m";
labels = {
severity = "medium";
severity = "critical";
};
annotations.summary = ''
Errors occur on {{ $labels.job }}
Disk {{ $labels.device }}
Disk {{ $labels.device }} {{ $value }}
{{ $labels.instance }}
'';
}
{
alert = "ZFSPoolsHealth";
expr = "zfs_pool_health > 0";
for = "5m";
labels = {
severity = "critical";
};
annotations.summary = ''
Unhealthy Pool at {{ $labels.job }}
Pool {{ $labels.pool }} value {{ $value }}
{{ $labels.instance }}
'';
}