Testing why a package won't deploy to ARM based lambda and to find a workaround.
I was seeing:
WARNING: The requested image's platform (linux/amd64) does not match the detected host platform (linux/arm64/v8) and no specific platform was requested
When running cdk deploy and could see it installing the x86 version of the package.
I was on a M2 (ARM) Mac and was specifying aws_lambda.Architecture.ARM_64 as my lambda arch.
But, here it is, compiling packages for x86! How disappointing.
This is mainly a CDK bootstrapped project.
The stack is defined in play_stack.stack and calls 2 constructs:
- The aws_lambda_python_alpha approach to building a python function.
- The turbolayers approach to building a python function.
Compare these 2 constrcuts to see how the approach differs.
Deploy and you can test them.
I have used https://github.com/CloudSnorkel/cdk-turbo-layers for some projects recently and this makes the deploy a lot faster by running the package installation on a remote lambda and then you access them all through a layer.
This means that you don't compile your packages on each deploy. Check the package docs for more.
It also means that even for locally triggered deploy I won't be building my dependancies locally.
And, that solved the issue - as well as bringing faster deploys.