知识卡片
扩容不能单独解除重试风暴
无限、无等待的重试会把一次依赖故障变成持续额外负载,甚至阻碍依赖扩容和重新启动。新节点若立即承受相同压力,也可能继续失效。控制流量和恢复余量必须先于或配合扩容,不能把重试视为无成本的可靠性手段。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
原书出处 · 1 处
- Resilience · 查看原书依据
even when the conditions that caused the downstream service to go down are resolved, it can’t come back up because it’s instantly brought under too much load. But, retries are a good thing, right? Yes, but whenever you implement retry logic, you should always include a backoff algorithm, which we’ll conveniently discuss in the next
参考来源
- 位置:《Cloud Native Go》第09章《Resilience》(来源项目book-sources;源文件:extracted/Cloud Native Go (Matthew A. Titmus)/_epub-src/OEBPS/ch09.xhtml)
- 结论依据:无限、无等待的重试会把一次依赖故障变成持续额外负载,甚至阻碍依赖扩容和重新启动。原文相应定义、示例或限制为此结论提供依据;补充纠错与版本边界在正文就近标明。
- 原始内容:even when the conditions that caused the downstream service to go down are resolved, it can’t come back up because it’s instantly brought under too much load. But, retries are a good thing, right? Yes, but whenever you implement retry logic, you should always include a backoff algorithm, which we’ll conveniently discuss in the next