Tracing shows how requests flow through distributed systems.
Concepts:
- Trace: The full path of a request
- Span: A single operation within a trace
- Context propagation: Passing trace IDs between services
Tools:
- Jaeger: CNCF project, popular open-source
- Zipkin: Original tracing system
- OpenTelemetry: Vendor-neutral standard (merges OpenTracing + OpenCensus)
When tracing helps:
- "Which service is causing latency?"
- "What's the call graph for this request?"
- "Where did this request fail?"
Interview tip: Know that tracing has overhead. Sample traces (-% of requests) in production.