知识卡片
故障隔离决定变更爆炸半径
模块虽可在代码中分开,若运行时共享故障域,一个报表模块的异常仍可能拖垮在线交易路径。隐藏依赖让局部变更的影响范围难以预测,使团队倾向整站停机以避免副作用。改善可用性应追踪并限制故障传播路径,而不只改变发布形式。
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。
原书出处
参考来源
- 位置:《Principles of Software Architecture Modernization》第1章《What’s Wrong with Monoliths?》(源文件:`_epub-src/OPS/ch01.xhtml`,归一化正文 [24977:29196))
- 结论依据:原文把隐藏联系和故障隔离不足联系起来,以报表模块拖垮在线系统为例,说明团队可能选择全应用停机以规避副作用。
- 原始内容:With monoliths and their high coupling, hidden ties make any change risky, inherent to the lack of failure isolation. Cascading failure is very dangerous. For example, your reporting module or analytics reports could tear down your online system, creating constant downtime. Your malfunctioning profile module could return broken data, corrupting your online reports. Often the safest route is to have full application downtime to avoid side effects.