Docker image to create Amazon Lambda packages

Why?

Some node and python modules require a build system because they contain c++ binding. Once deployed, the compiled module may not be compatible with the Amazon Lambda execution environment.

A common solution to this problem is to build the package on an EC2 instance using an Amazon Linux AMI and then to deploy it in Amazon Lambda. Building serverless applications, it is ironic to be obliged to use an EC2 server to deploy code.

This docker image is based on the Amazon Linux image and contains gcc.

The docker image myrmex/lambda-packager:1 contains python 2.7, python 3.6, python 3.7, pip, node 6.10, node 8.10 to create packages for Amazon Lambda.

The docker image for myrmex/lambda-packager:2 contains node 10.16 to create packages for Amazon Lambda.

Using the docker images from myrmex/lambda-packager, you can avoid errors like these during execution in Amazon Lambda:

/var/lang/lib/libstdc++.so.6: version `GLIBCXX_3.4.21' not found (required by /var/task/node_modules/bcrypt/lib/binding/bcrypt_lib.node)

Module version mismatch. Expected 46, got 48.

Usage

You can use a docker volume to mount the code of the Lambda in a container. The directory where npm install or pip install is executed inside the container is /data. By default, the installation will be performed for node.

docker run -v `pwd`:/data myrmex/lambda-packager:1

The image does not create the zip archive for your. It only installs the dependencies in an environment compatible with Lambda. You will still have to zip the result and deploy it in Amazon Lambda.

For a node package, take care that node_module does not already exist before running the command.

Building images locally

docker build -t myrmex/lambda-packager:1 -f al1.Dockerfile .
docker build -t myrmex/lambda-packager:2 -f al2.Dockerfile .

Managing permissions

The user account that performs npm install or pip install inside the container may have a uid/gid that differs from the uid/gid of the host machine and will not be able to perform the installation. The image myrmex/lambda-packager accepts two environment variables that allows to modify the uid/gid of the container's user: HOST_UID and HOST_GID. If HOST_GID is omitted, its value will be set to the value of HOST_UID.

docker run -e HOST_UID=`id -u` -e HOST_GID=`id -g` -v `pwd`:/data myrmex/lambda-packager

Selecting the runtime

The RUNTIME environment variable allows to choose the runtime.

Node 10.16

Node 10.16 is the default runtime for Amazon Linux 2 and does not require any special configuration.

docker run -v `pwd`:/data myrmex/lambda-packager:2

Node 8.10

Node 8.10 is the default runtime for Amazon Linux and does not require any special configuration.

docker run -v `pwd`:/data myrmex/lambda-packager:1

Node 6.10

docker run -e RUNTIME=node6 -v `pwd`:/data myrmex/lambda-packager:1

Python 3.7

Python 3.7 is the default Python 3 runtime and accepts the values python3 or python3.7

docker run -e RUNTIME=python3.7 -v `pwd`:/data myrmex/lambda-packager:1

Python 3.6

docker run -e RUNTIME=python3.6 -v `pwd`:/data myrmex/lambda-packager:1

Python 2.7

Accepts the values python or python2.

docker run -e RUNTIME=python -v `pwd`:/data myrmex/lambda-packager:1

Default command

To be able to change the user uid/gid, the container is executed with the root user. But to execute npm install, we need to use a non root user because it is a good practice and it avoids some bad surprises with node-gyp.

So, the default command is su myrmex -c "npm install --production", myrmex beeing the name of the non root user.

For python, the command is su myrmex -c "pip install --requirement requirements.txt --upgrade --target ."