Distributed tracing

Intent

Trace requests end-to-end across service boundaries using correlation IDs and spans.

Use when

• Always for microservices: you need to debug latency and failures across call graphs.
• You need to attribute work to downstream dependencies and isolate bottlenecks.

Avoid / watch-outs

• Don’t attach high-cardinality data (raw payloads, unbounded IDs) as span attributes.
• Sampling must preserve error traces; otherwise you lose the most important data.

Skill mapping

• observability: instrument root spans and downstream spans; define stable span names/attributes.
• resilience: correlate retries/breakers with trace attributes for debugging.