Building AWS layers

This is an example project for building custom AWS lambda layer combining pandas and numpy.

Prerequisites

You'll need Docker installed

Instructions

0. Fill in the specific requirements (with their versions) in the requirements.txt file.
1. Run:

cd ./aws_layer_builder
./get_layer_packages.sh
zip -r <some-meaningful-layer-name>.zip python

Behind the scenes: It installs the requirements.txt into a python folder. It generates a zip which you can now use to create AWS layer for your lambda function(s).

2. In the AWS console navigate to: Services > Lambda > Layers. Hit "Create Layer" and use the created zip. Copy the ARN.

3. Attach the layer to a lambda:

Option 1: Using the AWS console: Go to your lambda function. In the Designer section of the Configuration tab select Layers. Layers section appears below the Designer. Hit Add Layer and fill in the form.

or

Option 2: Using zappa, configure layers in the zappa_settings file. "layers": ["arn:aws:lambda:<region>:<account_id>:layer:<layer_name>:<layer_version>"] Note: Layers support is currently available here Miserlou/Zappa#1842

Notes:

Current example illustrates building a pandas+numpy layer for python3.6. The script uses Docker to get Lambda-compatible versions of the libraries listed in requirements.txt. It uses lambci/lambda:build-python3.6 (a public Docker image providing environment that replicates the live AWS Lambda environment almost identically.)

If you need a different environment, set IMAGE_NAME and PACKAGES_PATH accordingly (check get_layer_packages.sh).