Observability for Site Reliability Engineers
Observability is the ability to understand what a system is doing from the signals it emits: metrics, logs and traces. In SRE it supports monitoring distributed systems and practical alerting, such as the four golden signals of latency, traffic, errors and saturation. Good alerts page humans only for symptoms that need action.
Stage: Monitoring and Response · Part of the Site Reliability Engineer roadmap · Last reviewed 19 November 2026
What to learn in Observability
- Instrument a service with the four golden signals
- Tell symptom-based alerts from cause-based alerts and prefer symptoms
- Write alerts on SLO burn rate rather than raw thresholds
- Distinguish metrics, logs and traces and when each answers a question best
- Reduce alert noise by tuning thresholds, durations and routing
Observability interview questions
What are the four golden signals?
Latency, traffic, errors and saturation. Latency is the time to serve requests, traffic is demand, errors are failed requests, and saturation is how full a resource is. Together they give a compact first view of the health of almost any user-facing service.
Why are symptom-based alerts preferred over cause-based alerts?
Symptoms such as high error rate show real user impact and cover many possible causes. Cause alerts like high CPU often fire without any user harm, creating noise and fatigue. Causes are better kept as dashboards used during diagnosis.
Where Observability fits in the Site Reliability Engineer roadmap
Observability belongs to the Monitoring and Response stage (advanced level), which takes about 40 hours. See problems and fix them.
Other units in this stage
Back to the full Site Reliability Engineer roadmap