Зачем нужна распределённая трассировка

When one user request flows through many services, a failure or slowdown is hard to locate. Distributed tracing stitches the whole journey into a single, viewable timeline.

Two core concepts:

• A trace represents one end-to-end request
• A span represents one unit of work within it (e.g. a single service call)

Every trace gets a traceId; every span gets a spanId and a reference to its parent span. These IDs propagate from service to service so the pieces can be reassembled.

The trace context is carried between services, usually in HTTP headers (e.g. W3C traceparent). Downstream services read it and continue the same trace instead of starting a new one.

A trace view shows each span's duration as a waterfall, so you instantly see which service is the bottleneck, where an error originated, and how calls fan out.

Tracing complements the other two pillars:

• Metrics - aggregated numbers (latency, throughput)
• Logs - discrete events
• Traces - causal request flow across services

Micrometer Tracing is Spring's tracing abstraction. It creates and propagates spans for you, with pluggable backends (Brave/OpenTelemetry) and exporters (Zipkin).

Zipkin is a tracing backend: services send their finished spans to it, and it stores and visualizes the assembled traces in a web UI.

Much instrumentation is automatic: incoming HTTP requests, RestTemplate/WebClient calls, and messaging get spans without code changes once tracing is on.

Tracing every request can be expensive. Sampling records only a fraction of traces. A probability like 0.1 records 10%, balancing insight against overhead.

With tracing, debugging a slow or failing distributed request changes from guesswork across many logs to opening one trace and reading the waterfall. It is essential at scale.

Confirm the core tracing concepts.

You learned why tracing matters:

• A trace = whole request; a span = one unit of work
• traceId/spanId propagate via headers across services
• Micrometer creates spans; Zipkin stores and visualizes them
• Sampling controls overhead