Observability, Structured Logging & Health

Production systems must be observable without attaching a remote debugger:

1. Structured JSON Logging

• Never output unstructured string text (console.log("user logged in: " + user)).
• Log JSON payloads with standard keys:

  {
    "timestamp": "2026-09-13T21:40:00.000Z",
    "level": "info",
    "message": "User session authenticated",
    "service": "auth-service",
    "trace_id": "4bf92f3577b34da6a3ce929d0e0e4736",
    "span_id": "00f067aa0ba902b7",
    "user_id": "usr_98124",
    "duration_ms": 14.2
  }

2. Request Correlation & Context Propagation

• Extract or generate an x-request-id header at the API boundary.
• Propagate the request ID through async context and include it in all outbound HTTP/RPC client calls.

3. Production Health Probes

• Implement standardized health endpoints:
  • /healthz (Liveness): Returns 200 if the process event loop is responsive.
  • /readyz (Readiness): Returns 200 only if critical dependencies (database, message queue) are connected and ready to accept traffic.