Skip to content

Latest commit

 

History

History
403 lines (324 loc) · 23.7 KB

README.md

File metadata and controls

403 lines (324 loc) · 23.7 KB

AWS Message Processing Framework for .NET

nuget downloads build status

Notice: This library is in developer preview. It provides early access to upcoming features in the AWS Message Processing Framework for .NET. Any releases prior to 1.0.0 might include breaking changes.

The AWS Message Processing Framework for .NET is an AWS-native framework that simplifies the development of .NET message processing applications that use AWS services, such as Amazon Simple Queue Service (SQS), Amazon Simple Notification Service (SNS), and Amazon EventBridge. The framework reduces the amount of boiler-plate code developers need to write, allowing you to focus on your business logic when publishing and consuming messages.

  • For publishers, the framework serializes the message from a .NET object to a CloudEvents-compatible message, and then wraps that in the service-specific AWS message. It then publishes the message to the configured SQS queue, SNS topic, or EventBridge event bus.
  • For consumers, the framework deserializes the message to its .NET object and routes it to the appropriate business logic. The framework also keeps track of the message visibility while it is being processed (to avoid processing a message more than once), and deletes the message from the queue when completed. The framework supports consuming messages in both long-running polling processes and in AWS Lambda functions.

Project Status

The framework is currently in developer preview. The following features are supported:

  • Publishing to SQS, SNS, and EventBridge
  • Handling SQS messages in a long-running, polling process
  • Handling SQS messages in AWS Lambda functions
  • Handling messages from FIFO (first-in-first-out) queues, and respecting group ordering
  • OpenTelemetry Instrumentation
  • Customizing serialization
  • Performance and error hardening

Features to be added:

  • AWS X-Ray Instrumentation

Getting started

Add the AWS.Messaging NuGet package to your project:

dotnet add package AWS.Messaging --prerelease

The framework integrates with .NET's dependency injection (DI) service container. You can configure the framework during your application's startup by calling AddAWSMessageBus to add it to the DI container.

var builder = WebApplication.CreateBuilder(args);

// Register the AWS Message Processing Framework for .NET
builder.Services.AddAWSMessageBus(builder =>
{
    // Register that you'll publish messages of type ChatMessage to an existing queue
    builder.AddSQSPublisher<ChatMessage>("https://sqs.us-west-2.amazonaws.com/012345678910/MyAppProd");
});

The framework supports publishing one or more message types, processing one or more message types, or doing both in the same application.

Publishing Messages

The following code shows a configuration for an application that is publishing different message types to different AWS services.

var builder = WebApplication.CreateBuilder(args);

// Register the AWS Message Processing Framework for .NET
builder.Services.AddAWSMessageBus(builder =>
{
    // Register that you'll publish messages of type ChatMessage to an existing queue
    builder.AddSQSPublisher<ChatMessage>("https://sqs.us-west-2.amazonaws.com/012345678910/MyAppProd");

    // Register that you'll publish messages of type OrderInfo to an existing SNS topic
    builder.AddSNSPublisher<OrderInfo>("arn:aws:sns:us-west-2:012345678910:MyAppProd");

    // Register that you'll publish messages of type FoodItem to an existing EventBridge bus
    builder.AddEventBridgePublisher<FoodItem>("arn:aws:events:us-west-2:012345678910:event-bus/default");
});

Once you have registered the framework during startup, inject the generic IMessagePublisher into your code. Call its PublishAsync method to publish any of the message types that were configured above. The generic publisher will determine the destination to route the message to based on its type.

In the following example, an ASP.NET MVC controller receives both ChatMessage messages and OrderInfo events from users, and then publishes them to SQS and SNS respectively. Both message types can be published using the generic publisher that was configured above.

[ApiController]
[Route("[controller]")]
public class PublisherController : ControllerBase
{
    private readonly IMessagePublisher _messagePublisher;

    public PublisherController(IMessagePublisher messagePublisher)
    {
        _messagePublisher = messagePublisher;
    }

    [HttpPost("chatmessage", Name = "Chat Message")]
    public async Task<IActionResult> PublishChatMessage([FromBody] ChatMessage message)
    {
        // Perform business and validation logic on the ChatMessage here
        if (message == null)
        {
            return BadRequest("A chat message was not submitted. Unable to forward to the message queue.");
        }
        if (string.IsNullOrEmpty(message.MessageDescription))
        {
            return BadRequest("The MessageDescription cannot be null or empty.");
        }

        // Publish the ChatMessage to SQS, using the generic publisher
        await _messagePublisher.PublishAsync(message);

        return Ok();
    }

    [HttpPost("order", Name = "Order")]
    public async Task<IActionResult> PublishOrder([FromBody] OrderInfo message)
    {
        if (message == null)
        {
            return BadRequest("An order was not submitted.");
        }

        // Publish the OrderInfo to SNS, using the generic publisher
        await _messagePublisher.PublishAsync(message);

        return Ok();
    }
}

Service-specific publishers

The example shown above uses the generic IMessagePublisher, which can publish to any supported AWS service based on the configured message type. The framework also provides service-specific publishers for SQS, SNS and EventBridge. These specific publishers expose options that only apply to that service, and can be injected using the types ISQSPublisher, ISNSPublisher and IEventBridgePublisher.

For example, when publishing messages to an SQS FIFO queue, you must set the appropriate message group ID. The following code shows the ChatMessage example again, but now using an ISQSPublisher to set SQS-specific options.

public class PublisherController : ControllerBase
{
    private readonly ISQSPublisher _sqsPublisher;

    public PublisherController(ISQSPublisher sqsPublisher)
    {
        _sqsPublisher = sqsPublisher;
    }

    [HttpPost("chatmessage", Name = "Chat Message")]
    public async Task<IActionResult> PublishChatMessage([FromBody] ChatMessage message)
    {
        // Perform business and validation logic on the ChatMessage here
        if (message == null)
        {
            return BadRequest("A chat message was not submitted. Unable to forward to the message queue.");
        }
        if (string.IsNullOrEmpty(message.MessageDescription))
        {
            return BadRequest("The MessageDescription cannot be null or empty.");
        }

        // Send the ChatMessage to SQS using the injected ISQSPublisher, with SQS-specific options
        await _sqsPublisher.SendAsync(message, new SQSOptions
        {
            DelaySeconds = <delay-in-seconds>,
            MessageAttributes = <message-attributes>,
            MessageDeduplicationId = <message-deduplication-id>,
            MessageGroupId = <message-group-id>
        });

        return Ok();
    }
}

The same can be done for SNS and EventBridge, using ISNSPublisher and IEventBridgePublisher respectively.

await _snsPublisher.PublishAsync(message, new SNSOptions
{
    Subject = <subject>,
    MessageAttributes = <message-attributes>,
    MessageDeduplicationId = <message-deduplication-id>,
    MessageGroupId = <message-group-id>
});
await _eventBridgePublisher.PublishAsync(message, new EventBridgeOptions
{
    DetailType = <detail-type>,
    Resources = <resources>,
    Source = <source>,
    Time = <time>,
    TraceHeader = <trace-header>
});

Consuming Messages

To consume messages, implement a message handler using the IMessageHandler interface for each message type you wish to process. The mapping between message types and message handlers is configured in the project startup.

await Host.CreateDefaultBuilder(args)
    .ConfigureServices(services =>
    {
        // Register the AWS Message Processing Framework for .NET
        services.AddAWSMessageBus(builder =>
        {
            // Register an SQS Queue that the framework will poll for messages
            builder.AddSQSPoller("https://sqs.us-west-2.amazonaws.com/012345678910/MyAppProd");

            // Register all IMessageHandler implementations with the message type they should process. 
            // Here messages that match our ChatMessage .NET type will be handled by our ChatMessageHandler
            builder.AddMessageHandler<ChatMessageHandler, ChatMessage>();
        });
    })
    .Build()
    .RunAsync();

The following code shows a sample message handler for a ChatMessage message.

public class ChatMessageHandler : IMessageHandler<ChatMessage>
{
    public Task<MessageProcessStatus> HandleAsync(MessageEnvelope<ChatMessage> messageEnvelope, CancellationToken token = default)
    {
        // Add business and validation logic here
        if (messageEnvelope == null)
        {
            return Task.FromResult(MessageProcessStatus.Failed());
        }

        if (messageEnvelope.Message == null)
        {
            return Task.FromResult(MessageProcessStatus.Failed());
        }

        ChatMessage message = messageEnvelope.Message;

        Console.WriteLine($"Message Description: {message.MessageDescription}");

        // Return success so the framework will delete the message from the queue
        return Task.FromResult(MessageProcessStatus.Success());
    }
}

The outer MessageEnvelope contains metadata used by the framework. Its message property is the message type (in this case ChatMessage).

You can return MessageProcessStatus.Success() to indicate that the message was processed successfully and the framework will delete the message from the SQS queue. When returning MessageProcessStatus.Failed() the message will remain in the queue, where it can be processed again or moved to a dead-letter queue if configured.

Handling Messages in a Long-Running Process

You can call AddSQSPoller with an SQS queue URL to start a long-running BackgroundService that will continuously poll the queue and process messages.

await Host.CreateDefaultBuilder(args)
    .ConfigureServices(services =>
    {
        // Register the AWS Message Processing Framework for .NET
        services.AddAWSMessageBus(builder =>
        {
            // Register an SQS Queue that the framework will poll for messages
            builder.AddSQSPoller("https://sqs.us-west-2.amazonaws.com/012345678910/MyAppProd", options => 
            {
                // The maximum number of messages from this queue that the framework will process concurrently on this client
                options.MaxNumberOfConcurrentMessages = 10;

                // The duration each call to SQS will wait for new messages
                options.WaitTimeSeconds = 20; 
            });

            // Register all IMessageHandler implementations with the message type they should process
            builder.AddMessageHandler<ChatMessageHandler, ChatMessage>();
        });
    })
    .Build()
    .RunAsync();

Configuring the SQS Message Poller

The SQS message poller can be configured by the SQSMessagePollerOptions when calling AddSQSPoller.

  • MaxNumberOfConcurrentMessages - The maximum number of messages from the queue to process concurrently. The default value is 10.
  • WaitTimeSeconds - The duration (in seconds) for which the ReceiveMessage SQS call waits for a message to arrive in the queue before returning. If a message is available, the call returns sooner than WaitTimeSeconds. The default value is 20.

Message Visibility Timeout Handling

SQS messages have a visibility timeout period. When one consumer begins handling a given message, it remains in the queue but is hidden from other consumers to avoid processing it more than once. If the message is not handled and deleted before becoming visible again, another consumer may attempt to handle the same message.

The framework will track and attempt to extend the visibility timeout for messages that it is currently handling. You can configure this behavior on the SQSMessagePollerOptions when calling AddSQSPoller.

  • VisibilityTimeout - The duration in seconds that received messages are hidden from subsequent retrieve requests. The default value is 30.
  • VisibilityTimeoutExtensionThreshold - When a message's visibility timeout is within this many seconds of expiring, the framework will extend the visibility timeout (by another VisibilityTimeout seconds). The default value is 5.
  • VisibilityTimeoutExtensionHeartbeatInterval- How often in seconds that the framework will check for messages that are within VisibilityTimeoutExtensionThreshold seconds of expiring, and then extend their visibility timeout. The default value is 1.

In the following example the framework will check every 1 second for messages that are still being handled. For those messages within 5 seconds of becoming visible again, the framework will automatically extend the visibility timeout of each message by another 30 seconds.

 builder.AddSQSPoller("https://sqs.us-west-2.amazonaws.com/012345678910/MyAppProd", options => 
{
    options.VisibilityTimeout = 30;
    options.VisibilityTimeoutExtensionThreshold = 5;
    VisibilityTimeoutExtensionHeartbeatInterval = 1;
});

Handling Messages in AWS Lambda Functions

You can use the AWS Message Processing Framework for .NET with SQS's integration with Lambda. This is provided by the AWS.Messaging.Lambda package. Refer to its README to get started.

SQS Poller Resiliency

The SQS Poller is resilient by design and is able to handle errors thrown by the underlying .NET SDK as well as the framework itself in two ways. We have classified a number of exceptions that may occur due to invalid configuration from the user's side as fatal exceptions. These exceptions will cause the SQS Poller background service to stop running after throwing a user-friendly error message. However, any exceptions thrown, outside of the fatal ones we have defined, will not cause the SQS Poller to error out and will remain resilient in the event that an underlying service is facing degraded performance or outages. The SQS poller leverages backoffs in an effort to retry any failed SQS requests while applying a certain time-delay between retries.

The framework defined two interfaces, IBackoffHandler and IBackoffPolicy. The IBackoffPolicy is closely tied to the BackoffHandler which implements IBackoffHandler. The default implementation of IBackoffHandler checks the attached IBackoffPolicy to see if a backoff should be applied. If a backoff is to be applied, the IBackoffPolicy also returns the time delay between retries.

The framework support three backoff policies:

  • None, which would disable the backoff handler. This will allow users to fully rely on the SDK’s retry logic.
  • Interval, which would backoff on a given and configurable interval. Default value is 1 second.
  • CappedExponential, which would perform an exponential backoff up until it reaches a certain configurable max backoff time, at which point it would switch to an interval backoff. Default value for the cap backoff time is 1 hour.

By default, without any user configuration, the SQS Poller will use the default implementation of the IBackoffHandler interface, coupled with a capped exponential backoff policy.

Users are free to change the backoff policy as follows:

services.AddAWSMessageBus(builder =>
{
    builder.AddSQSPoller("https://sqs.us-west-2.amazonaws.com/536721586275/MPF");

    // Optional: Configure the backoff policy used by the SQS Poller.
    builder.ConfigureBackoffPolicy(options =>
    {
        // Use 1 of the available 3 backoff policies:

        // No backoff Policy
        options.UseNoBackoff();

        // Interval backoff policy
        options.UseIntervalBackoff(x =>
        {
            x.FixedInterval = 1;
        });

        // Capped exponential backoff policy
        options.UseCappedExponentialBackoff(x =>
        {
            x.CapBackoffTime = 60;
        });
    });
});

Users can also implement their own backoff handler by implementing the interface IBackoffHandler and injecting it before the AWS Message Bus is added to the DI Container. This can be done as follows:

services.TryAddSingleton<IBackoffHandler, CustomBackoffHandler>();

services.AddAWSMessageBus(builder =>
{
    builder.AddSQSPoller("https://sqs.us-west-2.amazonaws.com/012345678910/MPF");
    builder.AddMessageHandler<ChatMessageHandler, ChatMessage>("chatMessage");
});

As an example, you can use Polly to handle the retries. A sample app that achieves this could be found here.

Telemetry

The AWS Message Processing Framework for .NET is instrumented for OpenTelemetry to log traces for each message that is published or handled by the framework. This is provided by the AWS.Messaging.Telemetry.OpenTelemetry package. Refer to its README to get started.

Customization

The framework builds, sends, and handles messages in three different "layers":

  1. At the outermost layer, the framework builds the AWS-native request or response specific to a service. With SQS for example, it builds SendMessage requests, and works with the Message objects that are defined by the service.
  2. Inside the SQS request and response, it sets the MessageBody element (or Message for SNS or Detail for EventBridge) to a JSON-formatted CloudEvent. This contains metadata set by the framework that is accessible on the MessageEnvelope object when handling a message.
  3. At the innermost layer, the data attribute inside the CloudEvent JSON object contains a JSON serialization of the .NET object that was sent or received as the message.
{
    "id":"b02f156b-0f02-48cf-ae54-4fbbe05cffba",
    "source":"/aws/messaging",
    "specversion":"1.0",
    "type":"Publisher.Models.ChatMessage",
    "time":"2023-11-21T16:36:02.8957126+00:00",
    "data":"<the ChatMessage object serialized as JSON>"
}

You can customize how the message envelope is configured and read:

  • "id" uniquely identifies the message. By default it is set to a new GUID, but this can be overridden by implementing your own IMessageIdGenerator and injecting that into the DI container.
  • "type" controls how the message is routed to handlers. By default this uses the full name of the .NET type that corresponds to the message. You can override this via the messageTypeIdentifier parameter when mapping the message type to the destination via AddSQSPublisher, AddSNSPublisher, or AddEventBridgePublisher.
  • "source" indicates which system or server sent the message.
    • This will be the function name if publishing from AWS Lambda, the cluster name and task ARN if on Amazon ECS, the instance ID if on Amazon EC2, otherwise a fallback value of /aws/messaging.
    • You can override this via AddMessageSource or AddMessageSourceSuffix on the MessageBusBuilder.
  • "time" set to the current DateTime in UTC. This can be overridden by implementing your own IDateTimeHandler and injecting that into the DI container.
  • "data" contains a JSON representation of the .NET object that was sent or received as the message:
    • ConfigureSerializationOptions on MessageBusBuilder allows you to configure the System.Text.Json.JsonSerializerOptions that will be used when serializing and deserializing the message.
    • To inject additional attributes or transform the message envelope once the framework builds it, you can implement ISerializationCallback and register that via AddSerializationCallback on MessageBusBuilder.

Getting Help

For feature requests or issues using this framework please open an issue in this repository.

Contributing

We welcome community contributions and pull requests. See CONTRIBUTING.md for information on how to submit code.

Security

The AWS Message Processing Framework for .NET relies on the AWS SDK for .NET for communicating with AWS. Refer to the security section in the AWS SDK for .NET Developer Guide for more information. You can also find more information in AWS Message Processing Framework for .NET Developer Guide.

The framework does not log data messages sent by the user for security purposes. If users want to enable this functionality for debugging purposes, you need to call EnableMessageContentLogging() in the AWS Message Bus as follows:

builder.Services.AddAWSMessageBus(bus =>
{
    builder.EnableMessageContentLogging();
});

If you discover a potential security issue, refer to the security policy for reporting information.

Additional Resources

License

This project is licensed under the Apache-2.0 License.