Follow-ups to config generation reporting in symphony-server

未关闭
#77 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
4/5
预计耗时
3-5 天
新手友好度
35/100
Issue 类型
文档
描述清晰度
基本清楚
活跃度
活跃
技术栈
prometheus, rust

调研方向

Read issue #75 and the README sections describing config reloads, status.json, and the admin metrics endpoint. The issue lists five deferred edge cases but does not select one or define a change to make. Check whether a specific behavior should be revisited and what outcome is expected before estimating completion.

由索引模型根据 Issue 内容生成。

描述

These are small behaviors left as they are by Report the generation of the config a reload applied in status.json (#75). None of them gives a false appliedGeneration to a writer that reads status.json. Each is here so it can be revisited if someone runs into it.

  • reloadedAt advances on a reload that skipped a proxy. So does symphony_config_reload_time_seconds. This predates generations, and it stays because host-manager falls back to the timestamp on older symphony versions. Consider making it success-only once nothing relies on the old meaning.
  • /metrics.json and the generation samples describe the process that answered. During an overlapped upgrade the admin endpoint can answer from the outgoing process while status.json names the successor. The snapshot carries pid and the README says to compare it. Nothing ties the two together automatically.
  • Generation labels map CR to LF. escapeLabel normalizes line endings, so tokens that differ only by CR and LF render the same in Prometheus. status.json and /metrics.json keep the exact string, and writers use counters or hashes.
  • Two entries for one port-set where one fails. The port is listed in failedGeneration even if the other entry applied and now serves the new generation. host-manager writes one entry per port, so this needs a hand-written config.
  • A running proxy kept beside a successful split. A config that splits one proxy into two, where only one applies, keeps the old proxy because the skipped entry needs its port. The port that applied is then served by both proxies, so some connections get old routes, and status.json lists it twice. The README documents it. Removing the old proxy would leave the skipped port with nothing serving it.

Review-Deferral: a50f8cd314342795 severity=nit category=documentation
Review-Deferral: a4357533996f8c1b severity=nit category=documentation
Review-Deferral: aca0334a249e0dcb severity=nit category=correctness
Review-Deferral: 14972246fe8b3e0e severity=nit category=correctness
Review-Deferral: 40be5c430fabdc85 severity=minor category=correctness

🤖 Generated by Claude Opus 5.5; posted via @kriszyp.

主要语言
Rust
星标
1
派生
0
平均合并
19 小时 12 分钟
30 天内合并 PR
13

环境准备

  • 没有 Dockerfile 或 Docker Compose 文件
  • 没有 Pull Request 模板
  • 阅读贡献指南

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

HarperFast/symphony 的其他 Issue

查看 HarperFast/symphony 的全部 Issue

相似的 Issue

更多 Rust Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。