xpmatteo / spike-pandas-and-lambda Goto Github PK

An experiment to see if I can run Python packages that depend on native libraries on AWS Lambda.

Primary source of info: https://serverlesscode.com/post/deploy-scikitlearn-on-lamba/

Also helpful: https://alestic.com/2014/11/aws-lambda-cli/

Thanks to Alex Casalboni for pointing me in the right direction :-) https://www.quora.com/How-do-I-create-a-Python-AWS-Lambda-zip-with-OpenCV-+-FFmpeg/answer/Alex-Casalboni

Update: it is said that the only way to run native libraries in lambda is to compile them on the very same Linux image that Amazon provides. In the last commit, I proved that it also works with Ubuntu :) In previous commits, I did the same with the amazonlinux image.

How to use this repo

Check out what the Python code does in hello/main.py. It's basically just creating a Pandas data frame and printing it out.

Create the docker image with make build.

Then deploy the lambda function with make deploy. This will create the zip package and create or update the lambda function.

Invoke it with make invoke.

Check out the logs with make log. (Though the logs are probably easier to read on the AWS console.)