schema_version: 1
id: sql-tail-latency-spike-during-incident-window
title: SQL tail latency spike investigation
status: active
owner: claude-code
environment:
  profile: stable-premium
  cluster_id: '10748116073341772391'
  parent_id: '2026948324023603200'
  cluster_type: premium
  tidb_version: v8.5.4
time_window:
  incident_at: '2026-03-31T10:00:00Z'
  allowed:
  - id: workspace_observed_window
    start: '2026-03-31T10:00:00Z'
    end: '2026-03-31T12:11:00Z'
prompt:
  text: 请帮我分析 Premium 集群 10748116073341772391 在 2026-03-31T10:00:00Z 到 2026-03-31T12:11:00Z 这个时间窗内的 P999 SQL Duration 与尾延迟异常。业务侧反馈该时段 SQL 长尾延迟明显升高，部分 TiDB 实例 CPU 也偏高。请判断最可能的根因，并给出关键证据链、影响判断和处理建议。
tags:
- tail-latency
- p999-latency
- high-cpu
case_tier: P0
cost_tier: low
followup:
  scheduled_max_followups: 0
execution:
  acp_bash_timeout_s: 900