s3-balance/deploy/monitoring/s3_balance_alerts.yml

# S3 Balance 告警规则
groups:
  - name: s3_balance_alerts
    rules:
      # 存储桶不健康告警
      - alert: BucketUnhealthy
        expr: s3_balance_bucket_healthy == 0
        for: 5m
        labels:
          severity: warning
        annotations:
          summary: "存储桶 {{ $labels.bucket }} 不健康"
          description: "存储桶 {{ $labels.bucket }} ({{ $labels.endpoint }}) 已经连续5分钟不可用"

      # 错误率过高告警
      - alert: HighErrorRate
        expr: rate(s3_balance_s3_operations_total{status="error"}[5m]) > 0.1
        for: 2m
        labels:
          severity: critical
        annotations:
          summary: "存储桶 {{ $labels.bucket }} 错误率过高"
          description: "存储桶 {{ $labels.bucket }} 错误率超过10%，操作: {{ $labels.operation }}"

      # 操作延迟过高告警
      - alert: HighLatency
        expr: histogram_quantile(0.95, rate(s3_balance_s3_operation_duration_seconds_bucket[5m])) > 1
        for: 5m
        labels:
          severity: warning
        annotations:
          summary: "{{ $labels.operation }} 操作延迟过高"
          description: "{{ $labels.operation }} 操作95%分位数延迟超过1秒"

      # 存储桶使用率告警
      - alert: BucketUsageHigh
        expr: s3_balance_bucket_usage_bytes / s3_balance_bucket_capacity_bytes > 0.9
        for: 5m
        labels:
          severity: warning
        annotations:
          summary: "存储桶 {{ $labels.bucket }} 使用率过高"
          description: "存储桶 {{ $labels.bucket }} 使用率超过90%"

      # 存储桶满告警
      - alert: BucketFull
        expr: s3_balance_bucket_usage_bytes / s3_balance_bucket_capacity_bytes >= 0.95
        for: 1m
        labels:
          severity: critical
        annotations:
          summary: "存储桶 {{ $labels.bucket }} 即将满了"
          description: "存储桶 {{ $labels.bucket }} 使用率达到95%"

      # QPS过低告警 (可能服务异常)
      - alert: LowQPS
        expr: rate(s3_balance_s3_operations_total[5m]) == 0
        for: 10m
        labels:
          severity: warning
        annotations:
          summary: "QPS异常低"
          description: "系统QPS为0，可能服务异常无请求"

      # 负载均衡失败告警
      - alert: BalancerDecisionFailure
        expr: increase(s3_balance_balancer_decisions_total[5m]) == 0
        for: 5m
        labels:
          severity: warning
        annotations:
          summary: "负载均衡决策异常"
          description: "5分钟内没有负载均衡决策，可能存储桶全部不可用"