diff --git a/prometheus/alerts.yaml b/prometheus/alerts.yaml index b1e269b..947f5af 100644 --- a/prometheus/alerts.yaml +++ b/prometheus/alerts.yaml @@ -11,13 +11,13 @@ spec: - name: hardware rules: - alert: MemoryHigh - expr: round((((node_memory_MemTotal_bytes - node_memory_MemAvailable_bytes) / node_memory_MemTotal_bytes) * 100), 0.1) > 80 + expr: round((((node_memory_MemTotal_bytes - node_memory_MemAvailable_bytes) / node_memory_MemTotal_bytes) * 100), 0.1) > 90 for: 5m labels: severity: warning annotations: - summary: "Memory over 80%" - description: "Memory on node {{ $labels.node }} is over 80% for more than 5 minutes. Plox fix. Memory usage: {{ $value }}%" + summary: "Memory over 90%" + description: "Memory on node {{ $labels.node }} is over 90% for more than 5 minutes. Plox fix. Memory usage: {{ $value }}%" - alert: DiskspaceLowWorker expr: round(node_filesystem_avail_bytes{mountpoint="/", node=~"worker.*"} / 1073742000, 0.1) < 50 for: 1m