Quickstart: Host a Durable Task SDK app on Azure Container Apps (preview)

In this quickstart, you learn how to:

Prerequisites

Before you begin:

• Make sure you have .NET 8 SDK or later.
• Install Docker for running the emulator.
• Install Azure Developer CLI
• Clone the Durable Task Scheduler GitHub repository to use the quickstart sample.

• Make sure you have Python 3.9+ or later.
• Install Docker for running the emulator.
• Install Azure Developer CLI
• Clone the Durable Task Scheduler GitHub repository to use the quickstart sample.

Prepare the project

In a new terminal window, from the Azure-Samples/Durable-Task-Scheduler directory, navigate into the sample directory.

cd /samples/durable-task-sdks/dotnet/FunctionChaining

cd /samples/durable-task-sdks/python/function-chaining

Deploy using Azure Developer CLI

1. Run azd up to provision the infrastructure and deploy the application to Azure Container Apps in a single command.

   azd up
   

2. When prompted in the terminal, provide the following parameters.

   Parameter	Description
   Environment Name	Prefix for the resource group created to hold all Azure resources.
   Azure Location	The Azure location for your resources.
   Azure Subscription	The Azure subscription for your resources.

   This process may take some time to complete. As the azd up command completes, the CLI output displays two Azure portal links to monitor the deployment progress. The output also demonstrates how azd up:

   • Creates and configures all necessary Azure resources via the provided Bicep files in the ./infra directory using azd provision. Once provisioned by Azure Developer CLI, you can access these resources via the Azure portal. The files that provision the Azure resources include:
     • main.parameters.json
     • main.bicep
     • An app resources directory organized by functionality
     • A core reference library that contains the Bicep modules used by the azd template
   • Deploys the code using azd deploy

   Expected output

   Packaging services (azd package)
   
   (✓) Done: Packaging service client
   - Image Hash: {IMAGE_HASH}
   - Target Image: {TARGET_IMAGE}
   
   
   (✓) Done: Packaging service worker
   - Image Hash: {IMAGE_HASH}
   - Target Image: {TARGET_IMAGE}
   
   
   Provisioning Azure resources (azd provision)
   Provisioning Azure resources can take some time.
   
   Subscription: SUBSCRIPTION_NAME (SUBSCRIPTION_ID)
   Location: West US 2
   
    You can view detailed progress in the Azure Portal:
    https://portal.azure.com/#view/HubsExtension/DeploymentDetailsBlade/~/overview/id/%2Fsubscriptions%SUBSCRIPTION_ID%2Fproviders%2FMicrosoft.Resources%2Fdeployments%2FCONTAINER_APP_ENVIRONMENT
   
    (✓) Done: Resource group: GENERATED_RESOURCE_GROUP (1.385s)
    (✓) Done: Container Apps Environment: GENERATED_CONTAINER_APP_ENVIRONMENT (54.125s)
    (✓) Done: Container Registry: GENERATED_REGISTRY (1m27.747s)
    (✓) Done: Container App: SAMPLE_CLIENT_APP (21.39s)
    (✓) Done: Container App: SAMPLE_WORKER_APP (24.136s)   
   
   Deploying services (azd deploy)
   
    (✓) Done: Deploying service client
    - Endpoint: https://SAMPLE_CLIENT_APP.westus2.azurecontainerapps.io/
   
    (✓) Done: Deploying service worker
    - Endpoint: https://SAMPLE_WORKER_APP.westus2.azurecontainerapps.io/
   
   
   SUCCESS: Your up workflow to provision and deploy to Azure completed in 10 minutes 34 seconds.   
   

Confirm successful deployment

In the Azure portal, verify the orchestrations are running successfully.

1. Copy the resource group name from the terminal output.

2. Sign in to the Azure portal and search for that resource group name.

3. From the resource group overview page, click on the client container app resource.

4. Select Monitoring > Log stream.

5. Confirm the client container is logging the function chaining tasks.

   

6. Navigate back to the resource group page to select the worker container.

7. Select Monitoring > Log stream.

8. Confirm the worker container is logging the function chaining tasks.

   

Understanding the code

Client project

The Client project:

• Uses the same connection string logic as the worker
• Implements a sequential orchestration scheduler that:
  • Schedules 20 orchestration instances, one at a time
  • Waits 5 seconds between scheduling each orchestration
  • Tracks all orchestration instances in a list
  • Waits for all orchestrations to complete before exiting
• Uses standard logging to show progress and results

// Schedule 20 orchestrations sequentially
for (int i = 0; i < TotalOrchestrations; i++)
{
    // Create a unique instance ID
    string instanceName = $"{name}_{i+1}";

    // Schedule the orchestration
    string instanceId = await client.ScheduleNewOrchestrationInstanceAsync(
        "GreetingOrchestration", 
        instanceName);

    // Wait 5 seconds before scheduling the next one
    await Task.Delay(TimeSpan.FromSeconds(IntervalSeconds));
}

// Wait for all orchestrations to complete
foreach (string id in allInstanceIds)
{
    OrchestrationMetadata instance = await client.WaitForInstanceCompletionAsync(
        id, getInputsAndOutputs: false, CancellationToken.None);
}

Worker Project

The Worker project contains:

• GreetingOrchestration.cs: Defines the orchestrator and activity functions in a single file
• Program.cs: Sets up the worker host with proper connection string handling

Orchestration Implementation

The orchestration directly calls each activity in sequence using the standard CallActivityAsync method:

public override async Task<string> RunAsync(TaskOrchestrationContext context, string name)
{
    // Step 1: Say hello to the person
    string greeting = await context.CallActivityAsync<string>(nameof(SayHelloActivity), name);

    // Step 2: Process the greeting
    string processedGreeting = await context.CallActivityAsync<string>(nameof(ProcessGreetingActivity), greeting);

    // Step 3: Finalize the response
    string finalResponse = await context.CallActivityAsync<string>(nameof(FinalizeResponseActivity), processedGreeting);

    return finalResponse;
}

Each activity is implemented as a separate class decorated with the [DurableTask] attribute:

[DurableTask]
public class SayHelloActivity : TaskActivity<string, string>
{
    // Implementation details
}

The worker uses Microsoft.Extensions.Hosting for proper lifecycle management:

var builder = Host.CreateApplicationBuilder();
builder.Services.AddDurableTaskWorker()
    .AddTasks(registry => {
        registry.AddAllGeneratedTasks();
    })
    .UseDurableTaskScheduler(connectionString);
var host = builder.Build();
await host.StartAsync();

Client

The Client project:

• Uses the same connection string logic as the worker
• Implements a sequential orchestration scheduler that:
  • Schedules 20 orchestration instances, one at a time
  • Waits 5 seconds between scheduling each orchestration
  • Tracks all orchestration instances in a list
  • Waits for all orchestrations to complete before exiting
• Uses standard logging to show progress and results

# Schedule all orchestrations first
instance_ids = []
for i in range(TOTAL_ORCHESTRATIONS):
    try:
        # Create a unique instance name
        instance_name = f"{name}_{i+1}"
        logger.info(f"Scheduling orchestration #{i+1} ({instance_name})")

        # Schedule the orchestration
        instance_id = client.schedule_new_orchestration(
            "function_chaining_orchestrator",
            input=instance_name
        )

        instance_ids.append(instance_id)
        logger.info(f"Orchestration #{i+1} scheduled with ID: {instance_id}")

        # Wait before scheduling next orchestration (except for the last one)
        if i < TOTAL_ORCHESTRATIONS - 1:
            logger.info(f"Waiting {INTERVAL_SECONDS} seconds before scheduling next orchestration...")
        await asyncio.sleep(INTERVAL_SECONDS)
# ...
# Wait for all orchestrations to complete
for idx, instance_id in enumerate(instance_ids):
    try:
        logger.info(f"Waiting for orchestration {idx+1}/{len(instance_ids)} (ID: {instance_id})...")
        result = client.wait_for_orchestration_completion(
            instance_id,
            timeout=120
        )

Worker

Orchestration Implementation

The orchestration directly calls each activity in sequence using the standard call_activity function:

# Orchestrator function
def function_chaining_orchestrator(ctx, name: str) -> str:
    """Orchestrator that demonstrates function chaining pattern."""
    logger.info(f"Starting function chaining orchestration for {name}")

    # Call first activity - passing input directly without named parameter
    greeting = yield ctx.call_activity('say_hello', input=name)

    # Call second activity with the result from first activity
    processed_greeting = yield ctx.call_activity('process_greeting', input=greeting)

    # Call third activity with the result from second activity
    final_response = yield ctx.call_activity('finalize_response', input=processed_greeting)

    return final_response

Each activity is implemented as a separate function:

# Activity functions
def say_hello(ctx, name: str) -> str:
    """First activity that greets the user."""
    logger.info(f"Activity say_hello called with name: {name}")
    return f"Hello {name}!"

def process_greeting(ctx, greeting: str) -> str:
    """Second activity that processes the greeting."""
    logger.info(f"Activity process_greeting called with greeting: {greeting}")
    return f"{greeting} How are you today?"

def finalize_response(ctx, response: str) -> str:
    """Third activity that finalizes the response."""
    logger.info(f"Activity finalize_response called with response: {response}")
    return f"{response} I hope you're doing well!"

The worker uses DurableTaskSchedulerWorker for proper lifecycle management:

with DurableTaskSchedulerWorker(
    host_address=host_address, 
    secure_channel=endpoint != "http://localhost:8080",
    taskhub=taskhub_name, 
    token_credential=credential
) as worker:

    # Register activities and orchestrators
    worker.add_activity(say_hello)
    worker.add_activity(process_greeting)
    worker.add_activity(finalize_response)
    worker.add_orchestrator(function_chaining_orchestrator)

    # Start the worker (without awaiting)
    worker.start()