README.md

OpenTelemetry Distributed Tracing

Python | Durable Task SDK

Description

This sample demonstrates OpenTelemetry distributed tracing with the Durable Task SDK for Python. The SDK automatically creates activity spans with durabletask.* tags and propagates W3C trace context from orchestrations to activities — no manual span creation needed in your activity code.

Prerequisites

1. Python 3.9+
2. Docker

Quick Run

1. Start the infrastructure (emulator + Jaeger):

   docker compose up -d

2. Install dependencies and start the worker:

   python -m venv venv
   source venv/bin/activate  # Windows: venv\Scripts\activate
   pip install -r requirements.txt
   python worker.py

3. In a new terminal, run the client:

   source venv/bin/activate
   python client.py

4. View traces:

   • Jaeger UI: http://localhost:16686 (search for service DistributedTracingSample)
   • DTS Dashboard: http://localhost:8082

Viewing Traces

Open the Jaeger UI, select the DistributedTracingSample service, and click Find Traces. You'll see a single trace with 5 Spans — the parent create_orchestration:OrderProcessingOrchestration with 4 child activity spans:

Click on the trace to see the full span tree with parent-child hierarchy and rich durabletask.* tags (automatically added by the SDK):

How It Works

The Durable Task Python SDK has built-in OpenTelemetry support:

1. Client side: When you schedule an orchestration with an active OpenTelemetry span, the SDK captures the W3C trace context (traceparent/tracestate) and sends it with the request.
2. Orchestrator side: The SDK stores the parent trace context and passes it to all child activities and sub-orchestrations.
3. Activity side: The SDK wraps each activity execution in an activity:<name> span as a child of the orchestration's trace context, with durabletask.task.instance_id, durabletask.task.name, and durabletask.task.task_id tags.

All you need is to configure a TracerProvider with an exporter — the SDK does the rest.

What You'll See

The Jaeger UI shows a single trace for the entire orchestration with nested child spans for each activity. This helps you:

• Identify slow activities within an orchestration
• See the sequential flow of function chaining
• Correlate the full orchestration lifecycle in one trace
• Debug failures with full context

Clean Up

docker compose down

Learn More

• Observability Guide
• OpenTelemetry Python docs
• Durable Task Scheduler Dashboard