(aws-lambda): Ability to specify CPU architecture for building image

[SamStephens]

Describe the feature

I'd like the ability to specify usage of a bundling image for a particular CPU architecture. As an example, for a Java build, rather than saying:

        builder_options = BundlingOptions(
            command=packaging_instructions,
            image=aws_lambda.Runtime.JAVA_11.bundling_image,
            user="root",
            output_type=BundlingOutput.ARCHIVED,
        )

Which uses the architecture of the machine I'm building on, I'd like to be able to say something like:

        builder_options = BundlingOptions(
            command=packaging_instructions,
            image=aws_lambda.Runtime.JAVA_11.bundling_image_for(aws_lambda.Architecture.X86_64),
            user="root",
            output_type=BundlingOutput.ARCHIVED,
        )

Use Case

We use a Java library that doesn't work on ARM. We deploy to x86_64 Lambda, so this shouldn't be a problem. But in order to ensure builds on M1/M2 Mac work, we need to force the Docker image we use for Java to be an x86_64 image.

I'm currently using

        builder_options = BundlingOptions(
            command=packaging_instructions,
            image=aws_lambda.Runtime('java11:latest-x86_64', aws_lambda.RuntimeFamily.JAVA).bundling_image,
            user="root",
            output_type=BundlingOutput.ARCHIVED,
        )

As a workaround, but it'd be nice to be able to use the aws_lambda.Runtime and aws_lambda.Architecture constants.

I learned about this workaround from #18696 (comment)

Proposed Solution

No response

Other Information

No response

Acknowledgements

• I may be able to implement this feature request
• This feature might incur a breaking change

CDK version used

2.26.0, Python

Environment details (OS name and version, etc.)

Ubuntu 20.04 on WSL 2 within Windows 10.0.19044

[SamStephens]

Actually, my suggestion

aws_lambda.Runtime.JAVA_11.bundling_image_for(aws_lambda.Architecture.X86_64)

Is based on a Runtime being architecture independent. But the workaround

aws_lambda.Runtime('java11:latest-x86_64', aws_lambda.RuntimeFamily.JAVA).bundling_image

Implies that a Runtime can have an Architecture. Which is a little odd.

[HeskethGD]

Is it possible to specify the architecture on the bundling image in cdk? Think i'm hitting into this problem. Pydantic core is throwing errors in my deployment as it seems to not be matched up for the architecture and I think it may be related to this. I did swap everything to arm64 and had it working when using cdk deploy from my mac but now I'm deploying from github actions rather than my local machine it is breaking and i'm wondering if it is doing something different there that may be related to this. It suggests here that it is possible to specify architecture in bundling but I don't see it when i go to the actual BundlingOptions page https://docs.aws.amazon.com/cdk/api/v2/python/aws_cdk.aws_lambda_python_alpha/PythonLayerVersion.html

[tolikkostin]

@HeskethGD have you solved your issue? Seems I faced the same)

[SamStephens]

@HeskethGD @tolikkostin the workaround to specify an architecture is right in the original issue description. Rather than using an image constant like aws_lambda.Runtime.PYTHON_3_11.bundling_image, explicitly specify that you want the image for x86_64. In your case, something like

bundling_options = aws_lambda_python_alpha.BundlingOptions(
    image=aws_lambda.Runtime('python3.11:latest-x86_64', aws_lambda.RuntimeFamily.PYTHON).bundling_image,
)

Then use those bundling options as a parameter to PythonLayerVersion.

[tolikkostin]

@SamStephens Thanks that works for me.