Distributed tracing context propagation across HTTP and message queue boundaries

Traces break at service boundaries because trace context (trace ID, span ID, sampling decision) is not propagated in outgoing requests. Jaeger or Zipkin shows disconnected single-service traces instead of end-to-end request flows.

Use W3C Trace Context headers (traceparent, tracestate) for HTTP and inject/extract context for message queues.

HTTP (automatic with OTel auto-instrumentation):
The OTel HTTP instrumentation automatically injects traceparent into outgoing fetch/axios/http requests and extracts it from incoming requests. Verify with:

// Should see traceparent header in outgoing requests
console.log(outgoingRequest.headers['traceparent']);
// Example: 00-4bf92f3577b34da6a3ce929d0e0e4736-00f067aa0ba902b7-01

Message queues (manual):

const { context, propagation } = require('@opentelemetry/api');

// Producer: inject context into message headers
const headers = {};
propagation.inject(context.active(), headers);
await producer.send({ topic: 'orders', messages: [{ value: JSON.stringify(order), headers }] });

// Consumer: extract context from message headers
const parentContext = propagation.extract(context.active(), message.headers);
context.with(parentContext, () => {
  tracer.startActiveSpan('process-order', (span) => {
    // This span is now a child of the producer span
    span.end();
  });
});

Without context propagation, each service creates isolated root spans. The trace ID must flow with the request for the distributed trace to be stitched together in the backend.

Enabling end-to-end distributed tracing across HTTP services and/or message queues