Getting Started
Architecture
NServiceBus
Transports
Persistence
ServiceInsight
ServicePulse
ServiceControl
Monitoring
Samples

Using NServiceBus in AWS Lambda with SQS

Component: AWS Lambda (SQS)
NuGet Package: NServiceBus.AwsLambda.SQS (1.x)
Target Version: NServiceBus 8.x

This sample shows how to host NServiceBus within an AWS Lambda, in this case, a function triggered by incoming SQS messages. This enables hosting message handlers in AWS Lambda, gaining the abstraction of message handlers implemented using IHandleMessages<T> and also taking advantage of NServiceBus's extensible message-processing pipeline. This sample also shows a function triggered by an HTTP call and how to use NServiceBus to dispatch messages from within this context.

When hosting NServiceBus within AWS Lambda, the function handler class hosts an NServiceBus endpoint that is capable of processing multiple message types.

Prerequisites

The sample includes a CloudFormation template, which will deploy the Lambda and create the necessary queues.

The Amazon.Lambda.Tools CLI can be used to deploy the template to an AWS account.

  1. Install the Amazon.Lambda.Tools CLI
  2. Make sure an S3 bucket is available in the AWS region of choice
  3. Update the s3-bucket settings in aws-lambda-tools-defaults.json file found in the ServerlessEndpoint project with the name of the bucket
  4. Optionally change the stack-name setting

Running the sample

Run the following command from the ServerlessEndpoint directory to deploy the Lambda project:

dotnet lambda deploy-serverless

The deployment will ask for a stack name and an S3 bucket name to deploy the serverless stack.

After that, running the sample will launch a single console window:

  • RegularEndpoint is a console application that will send a TriggerMessage to the ServerlessEndpoint queue, which is monitored by the AWS Lambda.
  • The deployed ServerlessEndpoint project will receive messages from the ServerlessEndpoint queue and process them using the AWS Lambda runtime.

To try the AWS Lambda

  1. From the RegularEndpoint window, press Enter to send a TriggerMessage to the ServerLessEndpoint queue.
  2. The AWS Lambda will receive the TriggerMessage and hand off its processing to NServiceBus.
  3. The NServiceBus message handler for TriggerMessage on ServerlessEndpoint sends a ResponseMessage that will be handled by the RegularEndpoint

Code walk-through

The static NServiceBus endpoint must be configured using details that come from the AWS Lambda ILambdaContext. Since that is not available until a message is handled by the function, the NServiceBus endpoint instance is deferred until the first message is processed, using a lambda expression such as:

public static IAwsLambdaSQSEndpoint Configuration => new AwsLambdaSQSEndpoint(context =>
{
    var endpointConfiguration = new AwsLambdaSQSEndpointConfiguration("ServerlessEndpoint");
    endpointConfiguration.UseSerialization<SystemJsonSerializer>();

    var routing = endpointConfiguration.RoutingSettings;

    routing.RouteToEndpoint(typeof(TriggerMessage), "ServerlessEndpoint");
    routing.RouteToEndpoint(typeof(ResponseMessage), "RegularEndpoint");

    return endpointConfiguration;
});

The same class defines the AWS Lambda, which makes up the hosting for the NServiceBus endpoint. The FunctionHandler method hands-off processing of messages to NServiceBus:

public async Task FunctionHandler(SQSEvent evnt, ILambdaContext context)
{
    using var cancellationTokenSource =
        new CancellationTokenSource(context.RemainingTime.Subtract(DefaultRemainingTimeGracePeriod));

    await serverlessEndpoint.Process(evnt, context, cancellationTokenSource.Token);
}

Meanwhile, the message handler for TriggerMessage, also hosted within the AWS Lambda project, is a regular NServiceBus message handler that is capable of sending messages.

public class TriggerMessageHandler : IHandleMessages<TriggerMessage>
{
    static readonly ILog Log = LogManager.GetLogger<TriggerMessageHandler>();

    public async Task Handle(TriggerMessage message, IMessageHandlerContext context)
    {
        Log.Info($"Handling {nameof(TriggerMessage)} in ServerlessEndpoint.");
        await context.Send(new ResponseMessage());
    }
}

Dispatching a message outside of a message handler

There could be the need to dispatch a message after reacting to events other than messages being pushed to a queue. For example, responding to an S3 bucket file upload or to an HTTP request. This sample also demonstrates this use case.

  1. Open a browser and visit the URL produced during the execution of dotnet lambda deploy-serverless. The command produces a list of outputs; note the value for ApiURL.
  2. The AWS Lambda will receive the HTTP call and send a TriggerMessage to the ServerlessEndpoint queue.
  3. As in the previous example, the AWS Lambda will receive the TriggerMessage and hand off its processing to NServiceBus.
  4. The NServiceBus message handler for TriggerMessage on ServerlessEndpoint sends a ResponseMessage that will be handled by the RegularEndpoint
[LambdaFunction(Policies = "AWSLambda_FullAccess, AmazonSQSFullAccess")]
[HttpApi(LambdaHttpMethod.Get, "/")]
public async Task<string> HttpGetHandler(ILambdaContext context)
{
    await serverlessEndpoint.Send(new TriggerMessage(), context);
    return $"{nameof(TriggerMessage)} sent.";
}

Removing the sample stack

To remove the deployed stack enter:

dotnet lambda delete-serverless

and provide the previously chosen stack name.

Samples