Distributed tracing setup — correlating requests across services

In a microservices architecture, a single user request touches multiple services. When something fails or is slow, it's impossible to trace the request flow or identify which service caused the issue.

(1) Generate a trace ID at the entry point (API gateway or first service). (2) Propagate it via HTTP headers: W3C standard uses traceparent header. (3) Each service creates spans (units of work) linked to the trace. (4) Use OpenTelemetry SDK — it's the industry standard and supports all major languages. (5) Export traces to Jaeger, Zipkin, Datadog, or Honeycomb. (6) Auto-instrumentation: OpenTelemetry has libraries that automatically instrument HTTP clients, database drivers, and message queues. (7) Include trace ID in log messages for correlation. (8) Set sampling rate in production — tracing every request is expensive.

Logs show what happened in one service. Metrics show aggregate behavior. Traces show the complete journey of a single request across all services, with timing for each hop.