Code-Memo

Logging, metrics, and tracing

Logging

• Prefer structured logs for services (log/slog, zap, zerolog).
• Include request IDs / trace IDs, resource identifiers, and error context.
• Keep logs machine-parsable for indexing and alerting.

Metrics

• Export Prometheus metrics (or vendor format) for:
  • request rates/latency/error rates
  • queue depth / retries
  • Go runtime stats when useful (GC pauses, goroutines, heap)

Tracing

• Use OpenTelemetry when tracing helps correlate cross-service latency.
• Add spans around outbound calls (DB, HTTP, queue publish/consume).

Health checks

Expose endpoints or commands:

• liveness: process is alive
• readiness: dependencies available (DB connection, config loaded)

Correlation

Make it easy to connect:

• log line ↔ trace ↔ metrics label(s)
• stable identifiers (deployment, version, instance)