Summary
The four golden signals are latency, traffic, errors, and saturation. They are the minimum useful operational view of a service.
Interview Points
- Latency: how long requests take, including tail latency.
- Traffic: demand on the system, such as RPS or messages per second.
- Errors: failed requests, bad responses, exceptions, and rejected work.
- Saturation: how close resources are to exhaustion.
- Good alerting combines symptoms with actionable thresholds.
2-3 Minute Interview Script
“The four golden signals are a simple SRE framework for understanding service health: latency, traffic, errors, and saturation.
Latency tells me how long successful and failed requests take, especially p95 and p99. Traffic tells me how much demand the service is receiving. Errors tell me what percentage of work is failing. Saturation tells me whether a resource like CPU, memory, connections, threads, or queue depth is near its limit.
In an interview, I would use these signals to design monitoring for a critical API. I would alert on user-visible symptoms like high error rate or high tail latency, and use saturation metrics for diagnosis and capacity planning.
The senior-level point is that dashboards should answer two questions: are users impacted, and what is likely causing it?”
Follow-Ups
- Why is p99 latency important?
- Which metrics should page someone?