I am interest in running H2O not on an EC2 instance but from AWS Lambda.
The approach I took:
- Python runtime, and create a
deployment package .zip including the contents of <virtualenv>/lib/python2.7/site-packages/
- cp h2o.jar from S3 to /tmp/ in the init of the Lambda function, as it of course does not end up in <virtualenv>/lib/python2.7/site-packages/
- modified init() in h2o.py to pass jar_path='/tmp/h2o.jar' to H2OLocalServer.start(...
The jar is found but H2OLocalServer never properly starts:
"Server wasn't able to start in %f seconds."
Any guidance on how to fix this? And what is the most sensible approach?