Distributed tracing with OpenTelemetry
Follow one LLM request across gateway, retrieval, queue, prefill, decode and tools: traces, spans and context propagation, instrumenting a Python gateway with OpenTelemetry, span design for streaming, GenAI attributes, sampling strategies, exporting to Tempo or Jaeger, and linking traces to logs and metrics.
18 min·Verified 2026-11Pro+50 XP
A metric tells you the p95 TTFT rose. A log tells you that request 7b1f took 2.9 seconds. Neither tells you where those 2.9 seconds went. Distributed tracing does: it reconstructs the journey of a single request as a tree of timed operations across every service it touched.
Pro content
Keep reading with Inference Academy Pro
Unlock the rest of this lesson, plus every advanced lesson, hands-on lab, the larger playground models and completion certificates.