知识卡片
金丝雀需要分组观测
金丝雀让新旧版本并存,逐渐提高新版本流量,按版本与受众比较延迟、错误及业务结果。若只观察全局平均,小比例故障会被旧版成功掩盖;回退也需兼顾已写状态和在途请求。
这张卡片看不懂?
这张卡片的反馈已满 10 条,暂时无法继续提交。
已有反馈 · 0 / 10 条
还没有人提交反馈。
在整书笔记中的位置
原书出处 · 1 处
- Deployment Strategies and Production Readiness · 查看原书依据
The canary process begins by deploying the new version alongside the existing one. Initially, only a small percentage of traffic, perhaps 1% or 5%, routes to the new version. This limited exposure provides real production testing while minimizing potential impact.
参考来源
- 位置:《Building Distributed Systems Designing scalable architectures with microservices, event-driven patterns, and robust deployment…》第13章《Deployment Strategies and Production Readiness》(来源项目book-sources;源文件:extracted/Building Distributed Systems Designing scalable architectures with microservices, event-driven patterns, and robust deployment…/_epub-src/OPS/ch13.xhtml)
- 结论依据:原文的机制或对比支持以下核心判断:金丝雀让新旧版本并存,逐渐提高新版本流量,按版本与受众比较延迟、错误及业务结果。 本卡将相邻说明中的适用限制保留为解释。
- 原始内容:The canary process begins by deploying the new version alongside the existing one. Initially, only a small percentage of traffic, perhaps 1% or 5%, routes to the new version. This limited exposure provides real production testing while minimizing potential impact.