Structured Logging and Observability

Log levels:

• DEBUG: detailed diagnostic (off in production)
• INFO: normal operation milestones
• WARNING: recoverable unexpected state
• ERROR: operation failed, action needed

Structured logs (JSON) over free-form text:

import structlog
log = structlog.get_logger()
log.info("request_complete", method="POST", path="/api/data", status=200, latency_ms=42)

Metrics to expose: request rate, error rate, latency (p50/p95/p99), queue depth.

Health check endpoint: /health returning {"status": "ok"} — required for load balancers.

Anti-pattern: Logging only on error; you can't diagnose what you didn't observe.