Skip to content

swap 健康窗扩容:覆盖冷库 eager init 实测耗时 - #497

Merged
nghqqa merged 1 commit into
mainfrom
fix/swap-health-window-cold-db2
Oct 21, 2026
Merged

nghqqa merged 1 commit into
mainfrom
fix/swap-health-window-cold-db2

Conversation

@nghqqa

@nghqqa nghqqa commented Oct 21, 2026

Copy link
Copy Markdown
Owner

rc5 生产换版两连失败实证驱动。扩窗后重试将正常通过。详见 commit。

rc5 生产换版两连失败实证(2026-10-21):生产库(beta-mp-pg 刚重启+冷缓存)上,
新建容器的 mu eager schema init 需要 1-3 分钟(首轮 DNS 竞速 ENOTFOUND 一次+
冷库 DDL 慢),而主健康窗 12×5s(约 1 分钟)与回滚窗 6×1s 均短于该耗时——
rc5 健康门正确等待但窗口先耗尽 → 误回滚;且回滚容器同样处于 init 窗口期,
6×1s 判定过早触发 🛑 误报(实际回滚容器稍后即 mu:True healthy)。

- 主健康窗 12 轮 → 48 轮(5s 间隔不变,约 4 分钟)
- 回滚窗 6×1s → 36×2s(约 72 秒)
- 18700(小库热态)不受影响(2 轮即过);本修复由生产实测驱动,无语义变化

测试:release-tools 20/20(mock 首轮即 live,轮数扩容不影响用例时长)
@nghqqa
nghqqa merged commit 529b8ee into main Oct 21, 2026
5 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant