prometheus_rules_config_map.yaml

apiVersion: v1
kind: ConfigMap
metadata:
  creationTimestamp: null
  name: prometheus-rules
  namespace: monitoring
data:
  cpu-usage.rules: |
    ALERT NodeCPUUsage
      IF (100 - (avg by (instance) (irate(node_cpu{name="node-exporter",mode="idle"}[5m])) * 100)) > 5
      FOR 2m
      LABELS {
        severity="page"
      }
      ANNOTATIONS {
        SUMMARY = "机器{{$labels.instance}}: High CPU usage detected",
        DESCRIPTION = "机器事件{{$labels.instance}}: CPU usage is above 75% (current value is: {{ $value }})"
      }
  instance-availability.rules: |
    ALERT InstanceDown
      IF up == 0
      FOR 1m
      LABELS { severity = "page" }
      ANNOTATIONS {
        summary = "机器Instance {{ $labels.instance }} down了",
        description = "实例事件{{ $labels.instance }} of job {{ $labels.job }} has been down for more than 1 minute.",
      }
  low-disk-space.rules: |
    ALERT NodeLowRootDisk
      IF ((node_filesystem_size{mountpoint="/etc/hosts"} - node_filesystem_free{mountpoint="/etc/hosts"} ) / node_filesystem_size{mountpoint="/etc/hosts"} * 100) > 5
      FOR 2m
      LABELS {
        severity="page"
      }
      ANNOTATIONS {
        SUMMARY = "实例{{$labels.instance}}: Low root disk space",
        DESCRIPTION = "实例事件{{$labels.instance}}: Root disk usage is above 75% (current value is: {{ $value }})"
      }
 
    ALERT NodeLowDataDisk
      IF ((node_filesystem_size{mountpoint="/data-disk"} - node_filesystem_free{mountpoint="/data-disk"} ) / node_filesystem_size{mountpoint="/data-disk"} * 100) > 5
      FOR 2m
      LABELS {
        severity="page"
      }
      ANNOTATIONS {
        SUMMARY = "实例{{$labels.instance}}: Low data disk space",
        DESCRIPTION = "实例事件{{$labels.instance}}: Data disk usage is above 75% (current value is: {{ $value }})"
      }
  mem-usage.rules: |
    ALERT NodeSwapUsage
      IF (((node_memory_SwapTotal-node_memory_SwapFree)/node_memory_SwapTotal)*100) > 5
      FOR 2m
      LABELS {
        severity="page"
      }
      ANNOTATIONS {
        SUMMARY = "实例{{$labels.instance}}: Swap usage detected",
        DESCRIPTION = "实例事件{{$labels.instance}}: Swap usage usage is above 75% (current value is: {{ $value }})"
      }
 
    ALERT NodeMemoryUsage
      IF (((node_memory_MemTotal-node_memory_MemFree-node_memory_Cached)/(node_memory_MemTotal)*100)) > 5
      FOR 2m
      LABELS {
        severity="page"
      }
      ANNOTATIONS {
        SUMMARY = "实例{{$labels.instance}}: High memory usage detected",
        DESCRIPTION = "实例事件{{$labels.instance}}: Memory usage is above 75% (current value is: {{ $value }})"
      }