Azure Durable Functions: Fan-Out/Fan-In Pattern

Running a long process in a standard Azure Function hits the 5-10 minute timeout. Durable Functions allows you to write stateful workflows in code. The most powerful pattern is Fan-Out/Fan-In.

The Scenario

You need to resize 10,000 images uploaded to Blob Storage.

[FunctionName("Orchestrator")]
public static async Task RunOrchestrator(
    [OrchestrationTrigger] IDurableOrchestrationContext context)
{
    var blobs = await context.CallActivityAsync<List<string>>("GetBlobList", null);
    
    var tasks = new List<Task<string>>();
    foreach (var blob in blobs)
    {
        // Fan-Out: This schedules 10,000 functions in parallel!
        tasks.Add(context.CallActivityAsync<string>("ResizeImage", blob));
    }

    // Fan-In: Wait for all to finish
    await Task.WhenAll(tasks);
    
    // Aggregation
    await context.CallActivityAsync("SendReport", tasks.Select(t => t.Result).ToList());
}

Behind the scenes, the Durable Task Framework manages the Azure Storage Queues to track the state of every single function execution. It creates a robust checkpointing system automatically.