alerts.yml 2.0 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445
  1. groups:
  2. - name: zhifu
  3. rules:
  4. - alert: ServiceUnavailable
  5. expr: up{job=~"platform|inference|operations|replica-gateway"} == 0
  6. for: 30s
  7. labels: {severity: critical}
  8. annotations:
  9. summary: "服务监控采集失败"
  10. description: "检查目标进程、网络及依赖服务。"
  11. - alert: WorkerUnavailable
  12. expr: zhifu_workers_alive == 0
  13. for: 15s
  14. labels: {severity: critical}
  15. annotations:
  16. summary: "工作进程心跳中断"
  17. description: "检查对应 role 的进程日志并恢复服务。"
  18. - alert: QueueWaiting
  19. expr: zhifu_oldest_pending_seconds > 120
  20. for: 1m
  21. labels: {severity: warning}
  22. annotations:
  23. summary: "任务等待超过两分钟"
  24. description: "检查模型额度、工作进程及外部业务响应。"
  25. - alert: HttpServerErrors
  26. expr: sum by(job) (rate(http_requests_total{status=~"5.."}[5m])) / clamp_min(sum by(job) (rate(http_requests_total[5m])), 0.01) > 0.05
  27. for: 1m
  28. labels: {severity: critical}
  29. annotations:
  30. summary: "HTTP 服务错误比例超过百分之五"
  31. description: "依据请求编号检查应用日志及依赖状态。"
  32. - alert: InferenceSaturation
  33. expr: sum(rate(http_requests_total{job="inference",status="429"}[5m])) / clamp_min(sum(rate(http_requests_total{job="inference"}[5m])), 0.01) > 0.1
  34. for: 1m
  35. labels: {severity: warning}
  36. annotations:
  37. summary: "向量服务繁忙请求超过百分之十"
  38. description: "检查计算资源、请求批量和服务副本数量。"
  39. - alert: AgentFailures
  40. expr: zhifu_jobs_15m{kind="agent",status="FAILED"} >= 5
  41. for: 1m
  42. labels: {severity: warning}
  43. annotations:
  44. summary: "最近十五分钟 Agent 失败任务达到五项"
  45. description: "查询任务 error_code 和模型服务状态。"