Add nats host to monitoring

2025-02-11 23:12:55 +01:00 · 2025-02-11 23:12:55 +01:00 · f0bc29ac5e
commit f0bc29ac5e
parent 539ff4eeac
1 changed files with 20 additions and 9 deletions
--- a/services/monitoring/prometheus.nix
+++ b/services/monitoring/prometheus.nix
@ -95,6 +95,16 @@
                annotations:
                  summary: "Caddy not running on {{ $labels.instance }}"
                  description: "Caddy has been down on {{ $labels.instance }} more than 5 minutes."
+          - name: nats_rules
+            rules:
+              - alert: nats down
+                expr: node_systemd_unit_state {instance="nats1.home.2rjus.net:9100", name = "nats.service", state = "active"} == 0
+                for: 5m
+                labels:
+                  severity: critical
+                annotations:
+                  summary: "NATS not running on {{ $labels.instance }}"
+                  description: "NATS has been down on {{ $labels.instance }} more than 5 minutes."
          - name: home_assistant_rules 
            rules:
              - alert: home_assistant_down 
@ -140,6 +150,7 @@
              "ns1.home.2rjus.net:9100"
              "ns2.home.2rjus.net:9100"
              "pgdb1.home.2rjus.net:9100"
+              "nats1.home.2rjus.net:9100"
            ];
          }
        ];