intermediate · slos

What Should Page Someone at 3am?

What deserves to wake a human at 3am, and what does not?

Practical scenario

An on-call rotation receives roughly 40 alerts a week. Two of them last quarter corresponded to real user-facing incidents. Nobody wants to be on call.

What it tests

  • Symptom-based versus cause-based alerting
  • Whether the candidate requires actionability
  • Awareness of alert fatigue as a real failure mode

Lessons behind this question