Jorge Dias

Recently I’ve been using AWS Lambda at work for some projects and one of the limitations that you have is trying to use packages with compiled dependencies.

Compiling them on your machine won’t work and the “recommended” way is to start an EC2 instance and compile your dependencies there and then copy those to your machine back.

Fortunately AWS now provides a docker image for amazon linux which we can leverage to build our depencies. We can avoid launching an instance and get faster results.

Below I provide an example to build the dependencies for a python project but you can probably modify it for other cases.

Just put the following content in a Makefile.

compile:
	yum install -y gcc libffi-devel openssl-devel python27-virtualenv
	virtualenv /tmp/venv
	/tmp/venv/bin/pip install --upgrade pip setuptools
	/tmp/venv/bin/pip install -e .
	cp -r /tmp/venv/lib/python2.7/site-packages/. ./aws_lambda_libs
	cp -r /tmp/venv/lib64/python2.7/site-packages/. ./aws_lambda_libs

lambda-deps:
	@echo "--> Compiling lambda dependencies"
	docker run --rm -it -v ${CURDIR}:/src -w /src amazonlinux make compile

.PHONY: lambda-deps compile

Then you just need to execute:

make lambda-deps

When this process finishes you’ll have your dependencies compiled in aws_lambda_libs.

Just include those contents at the root of your package.zip and deploy.

Enjoy!