This sample demonstrates distributed tracing with the Durable Task Java SDK using OpenTelemetry. Traces are exported to Jaeger via OTLP/gRPC for visualization.

The Java SDK automatically propagates W3C trace context (traceparent/tracestate) when scheduling orchestrations. The worker creates spans around each activity and orchestration execution, forming a correlated trace tree visible in Jaeger.

• Java 17+
• Docker (for DTS emulator and Jaeger)

# Start the DTS emulator (port 8080 for gRPC, 8082 for dashboard)
docker run -d --name dts-emulator \
  -p 8080:8080 -p 8082:8082 \
  mcr.microsoft.com/dts/dts-emulator:latest

# Start Jaeger (port 16686 for UI, 4317 for OTLP gRPC)
docker run -d --name jaeger \
  -p 16686:16686 -p 4317:4317 -p 4318:4318 \
  -e COLLECTOR_OTLP_ENABLED=true \
  jaegertracing/all-in-one:latest

./gradlew :samples:runTracingPattern -PskipSigning -x downloadProtoFiles

• Jaeger UI: http://localhost:16686 — Search for service durabletask-java-tracing-sample
• DTS Dashboard: http://localhost:8082

The TracingPattern sample demonstrates the Fan-Out/Fan-In pattern with distributed tracing:

1. Configures OpenTelemetry with an OTLP exporter pointing to Jaeger
2. Connects a worker and client to the DTS emulator using a connection string
3. Creates a parent span (create_orchestration:FanOutFanIn) and schedules an orchestration
4. The orchestration waits on a 1-second durable timer, then fans out 5 parallel GetWeather activities (Seattle, Tokyo, London, Paris, Sydney), fans in the results, then calls CreateSummary to aggregate
5. The SDK automatically propagates trace context through the full execution chain

Shows the trace from durabletask-java-tracing-sample service with spans covering the full fan-out/fan-in orchestration lifecycle.

Full span hierarchy showing the fan-out/fan-in pattern with proper span durations:

• create_orchestration:FanOutFanIn (root, internal)
  • orchestration:FanOutFanIn (server — full lifecycle, ~1.2s)
  • orchestration:FanOutFanIn:timer (internal — creation-to-fired, ~965ms)
  • activity:GetWeather ×5 (client — scheduling-to-completion, ~184ms) + ×5 (server — execution, ~25ms)
  • activity:CreateSummary (client, ~8ms) + (server, ~0.7ms)

15 spans total, Depth 2.

Note: Java OTel doesn't support SetSpanId() like .NET, so child spans appear as siblings under create_orchestration rather than nested under the orchestration span. All spans have meaningful durations.

Activity span showing attributes aligned with the .NET SDK schema:

• durabletask.type=activity
• durabletask.task.name=GetWeather
• durabletask.task.instance_id=<orchestrationId>
• durabletask.task.task_id=1
• otel.scope.name=Microsoft.DurableTask
• span.kind=server

The FanOutFanIn orchestration completed successfully with all activities.

docker stop jaeger dts-emulator && docker rm jaeger dts-emulator

The samples-azure-functions module contains a matching Fan-Out/Fan-In sample (TracingChain.java) for use with Azure Durable Functions. It uses the same pattern (1s timer → 5× GetWeather → CreateSummary) but runs as an HTTP-triggered Azure Function.

cd samples-azure-functions
../gradlew azureFunctionsPackage -PskipSigning -x downloadProtoFiles
cd build/azure-functions/<app-name>
func start

Then trigger with: curl http://localhost:7071/api/StartFanOutFanIn

Distributed tracing in Durable Functions exports to Application Insights (not Jaeger/OTLP). Configure your APPLICATIONINSIGHTS_CONNECTION_STRING in local.settings.json to see traces.