知识卡片
恢复要停止过时工作
短故障可能积累大量已无人等待的工作,即使依赖恢复,系统仍花很久追赶。端到端截止时间与取消可停止失效请求,按用户时延扩容可加速消化积压,适时优先较新任务可恢复交互体验。舍弃旧工作只适合允许失效的任务,账务等必须完成的工作不能照搬。
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
- Chapter 16. Common Failure Patterns · 查看原书依据
Every request that enters the system should have a timeout, the time after which the request is considered invalid and failed. When the request times out, any work associated with that request is aborted. Timeouts can help the system shed work when significantly overloaded. However, it is important
参考来源
- 位置:《Designing Distributed Systems, 2nd Edition》第16章《Chapter 16. Common Failure Patterns》(来源项目 book-sources;源文件:extracted/Designing Distributed Systems, 2nd Edition/_epub-src/OEBPS/ch16.html)
- 结论依据:原文提出超时、按时延扩容和优先新请求三种积压恢复手段。
- 原始内容:Every request that enters the system should have a timeout, the time after which the request is considered invalid and failed. When the request times out, any work associated with that request is aborted. Timeouts can help the system shed work when significantly overloaded. However, it is important