sqs-lambda | Rust crate for writing AWS Lambdas | AWS library

You are creating a Role to run your lambda, a lambda function, and permissions for something to invoke that lambda. The SourceArn is the thing that will invoke the lambda. So in your case it sounds like you want an S3 bucket to invoke the lambda, so the SourceArn would be the ARN of the S3 bucket.

This tutorial relates to what you are doing--specifically step 8 under the "Create the Lambda function" section.

Your CF template generally looks correct. The only thing I see that will be assuming this role is lambda.amazonaws.com, so the role may not need to list the following in the AssumeRolePolicyDocument section:

both refer to ECS Services that I'm not using due to the Desired Count variable which will make always a task run and this will lead to an unnecessary cost.

ECS Services is exactly what you need to use here. Desired count can be 0 when there are no messages in the queue, which will prevent any charges from being incurred. You should create an ECS service that polls the SQS queue, and configure the ECS auto-scaling to scale from 0 to N service instances based on the available message count in the queue.

Found answer it was mistake only.

As per explaination given in AWS Lambda w/ SQS trigger, SQS Lambda Destinations, never adds to destination queue is correct only for calling lambda async you have to run different command. And test input from UI doesn't work like async. It works as sync call.

Yours and @JasonWadsworth's conclusion is absolutely correct. It's quite straightforward actually.

• Lambda functions "triggered" by an SQS message event are invoked synchronously, as per the documentation:

Lambda polls the queue and invokes your function synchronously with an event that contains queue messages.

• Destinations only works when the lambda function is invoked asynchronously, as alluded to in your link:

... AWS Lambda Destinations for asynchronous invocations.

(Although at no point does it explicitly state that synchronous invocations are not supported, a quick google will verify that from other sources.)

• You invoked your function from the CLI asynchronously.

--invocation-type Event

As per the documentation:

To invoke a function asynchronously, set InvocationType to Event.

Therefore what happened was:

1. Lambda(foo) was invoked asynchronously. Because the invocation was asynchronous,
2. The function result was written to Queue(Alpha). Because you've set Lambda(bar) as a lambda trigger on this queue, and because Lambda invokes functions synchronously based on SQS messages,
3. Lambda(bar) was invoked synchronously. Because it was invoked synchronously,
4. Queue(Beta) was not written to, and the pipeline stopped there.

We have a very similar pipeline with lambda functions and queues, the only difference is our functions write to the next queue in the pipeline using the SQS sdk. Not sure if it's the best way, but it's certainly one way to do it.