schema_version: 1
id: checkpoint-catchup-slow-after-changefeed-resume
title: Checkpoint catch-up is slow after changefeed resume
status: draft
owner: codex
source: jira
environment:
  profile: stable-dedicated
  cluster_id: '10119758836306799399'
  cluster_type: dedicated
  tidb_version: v8.5.7-release.1
time_window:
  incident_at: '2026-07-23T02:53:00Z'
  allowed:
  - id: reported_incident_window
    start: '2026-07-23T02:40:00Z'
    end: '2026-07-23T03:05:00Z'
prompt:
  text: 请分析集群 10119758836306799399 在 2026-07-23T02:40:00Z 到 2026-07-23T03:05:00Z 的 changefeed 恢复问题。changefeed 在约 02:53Z 重启后 resolved-ts 很快恢复，但 checkpoint-ts 每个 flush 周期只推进约 2 秒，积压无法正常追平。请判断瓶颈在哪里、哪些配置会放大问题，并给出恢复建议。
tags:
- cdc
- replication-lag
- recovery
case_tier: P0
cost_tier: medium
followup:
  scheduled_max_followups: 0
execution:
  acp_bash_timeout_s: 900