Observability

Observability makes runtime behavior visible enough to debug quickly.

Three pillars to implement

• logs
• metrics
• traces (or at minimum request IDs)

Logging baseline

• structured logs in production
• request IDs in every app log line
• access logs enabled where auditability is required

Metrics baseline

Track at least:

• request rate
• error rate
• p50/p95/p99 latency
• active connections
• worker restarts

Tracing/request correlation

If full tracing is unavailable, add request ID middleware and propagate ID in headers/logs.

Alerting priorities

• sustained error-rate increase
• latency SLO violation
• repeated worker crash loops
• health endpoint failures

Non-technical summary

Observability is how teams answer three incident questions quickly:

• what is broken
• how bad it is
• where to act first