← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 8790.c · 生成模型:GPT-6

监控覆盖测试未见过的情形

理解等级:中级

有限测试不能穷尽未来输入和故障,生产监控负责发现未知场景、漂移与退化。结合服务、数据、模型和业务信号,才能避免只看接口返回正常而漏掉静默误预测。发现问题后还要形成可执行的处置和测试补充回路。

这张卡片看不懂?

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

    • · 查看原书依据

      Monitoring gives us insight into scenarios that we haven’t considered before in our tests. As Edsger W. Dijkstra once said: “Testing may convincingly demonstrate the presence of bugs but can never demonstrate their absence.” That’s why monitoring in production is an essential complementary practice to testing.

    参考来源

    - 位置:《Effective Machine Learning Teams (David Tan, Ada Leung, David Colls)》第09章《MLOps与持续交付》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Effective Machine Learning Teams (David Tan, Ada Leung, David Colls)/_epub-src/OEBPS/ch09.html) - 结论依据:摘录中关于监控覆盖测试未见过的情形的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。 - 原始内容:Monitoring gives us insight into scenarios that we haven’t considered before in our tests. As Edsger W. Dijkstra once said: “Testing may convincingly demonstrate the presence of bugs but can never demonstrate their absence.” That’s why monitoring in production is an essential complementary practice to testing.