schema_version: 1
id: running-changefeed-with-stalled-progress
title: Running changefeed with stalled progress
status: draft
owner: codex
source: jira
environment:
  profile: stable-dedicated
  cluster_id: '1379661944646413949'
  cluster_type: dedicated
  tidb_version: v8.5.3
time_window:
  incident_at: '2026-07-13T12:00:00Z'
  allowed:
  - id: reported_incident_window
    start: '2026-07-13T11:50:00Z'
    end: '2026-07-13T12:20:00Z'
prompt:
  text: 请分析集群 1379661944646413949 在 2026-07-13T11:50:00Z 到 2026-07-13T12:20:00Z 的 changefeed 异常。changefeed 状态仍显示 Running，但 checkpoint 长时间不再推进；日志在异常前后出现 Kafka broken pipe、EOF 或 connection reset，Pause/Resume 后恢复。请判断根因和长期处理方式。
tags:
- cdc
- replication-lag
- kafka
case_tier: P0
cost_tier: medium
followup:
  scheduled_max_followups: 0
execution:
  acp_bash_timeout_s: 900