知识卡片
背压把过载传回源头
下游过载时返回明确错误,如服务不可用,上游继续把压力沿调用路径传播,让客户端减速、重试或放弃。没有背压,消息在服务内堆积,可能把局部拥堵扩大成内存耗尽。重试仍须受控,不能无条件立即重发。节点级保护处理分配不均,集群负载入口还要按当前有效实例能力调总上限,两个层级互相补充。
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
- 构建稳定系统 · 查看原书依据
When a downstream service in a cascade of services is experiencing a load spike, the service must propagate an error back to the services before them. Typically, this is done by sending an error or HTTP code such as 503 – Service Unavailable.
参考来源
- 位置:《Software Architecture and Decision-Making》第12章《Building Stable Systems》(来源项目:book-sources;源文件:extracted/Software Architecture and Decision-Making Leveraging Leadership, Technology, and Product Management to Build Great Products… (Srinath Perera)/_epub-src/ch12.xhtml)
- 结论依据:摘录说明下游过载时返回明确错误,如服务不可用,上游继续把压力沿调用路径传播,让客户端减速、重试或放弃。其余条件结合本章相邻例子与取舍说明,数字示例不作当前性能承诺。
- 原始内容:When a downstream service in a cascade of services is experiencing a load spike, the service must propagate an error back to the services before them. Typically, this is done by sending an error or HTTP code such as 503 – Service Unavailable.