prometheus: monitor link usage for switches

group_vars/prom/prometheus/switch.yml

@@ -44,4 +44,48 @@ prometheus__rules_switch:
       ErrorRate: !unsafe '{{ printf "%.2f" $value }} %'
       Interface: !unsafe "{{ $labels.ifName }}
         {{ if $labels.ifAlias }}- {{ $labels.ifAlias }}{{ end }}"
+  - alert: SwitchInLinkUsage
+    expr:
+      100 * rate(ifHCInOctets[5m]) / (ifHighSpeed * 1000000 / 8) > 50
+    for: 5m
+    keep_firing_for: 10m
+    labels:
+      severity: warning
+    annotations:
+      Usage: !unsafe '{{ printf "%.2f" $value }} %'
+      Interface: !unsafe "{{ $labels.ifName }}
+        {{ if $labels.ifAlias }}- {{ $labels.ifAlias }}{{ end }}"
+  - alert: SwitchInLinkUsage
+    expr:
+      100 * rate(ifHCInOctets[5m]) / (ifHighSpeed * 1000000 / 8) > 80
+    for: 5m
+    keep_firing_for: 10m
+    labels:
+      severity: critical
+    annotations:
+      Usage: !unsafe '{{ printf "%.2f" $value }} %'
+      Interface: !unsafe "{{ $labels.ifName }}
+        {{ if $labels.ifAlias }}- {{ $labels.ifAlias }}{{ end }}"
+  - alert: SwitchOutLinkUsage
+    expr:
+      100 * rate(ifHCOutOctets[5m]) / (ifHighSpeed * 1000000 / 8) > 50
+    for: 5m
+    keep_firing_for: 10m
+    labels:
+      severity: warning
+    annotations:
+      Usage: !unsafe '{{ printf "%.2f" $value }} %'
+      Interface: !unsafe "{{ $labels.ifName }}
+        {{ if $labels.ifAlias }}- {{ $labels.ifAlias }}{{ end }}"
+  - alert: SwitchOutLinkUsage
+    expr:
+      100 * rate(ifHCOutOctets[5m]) / (ifHighSpeed * 1000000 / 8) > 80
+    for: 5m
+    keep_firing_for: 10m
+    labels:
+      severity: warning
+    annotations:
+      Usage: !unsafe '{{ printf "%.2f" $value }} %'
+      Interface: !unsafe "{{ $labels.ifName }}
+        {{ if $labels.ifAlias }}- {{ $labels.ifAlias }}{{ end }}"
 ...