Interactive notebooks for trying analyses and exploring datasets

We're using the tmpnb service to launch docker containers on demand. Each user who launches from notebooks.codeneuro.org gets their own temporary interactive environment, preloaded with a variety of executable notebooks. The environments will be deleted after an hour of inactivity, so this is not intended for real work! But it is a great way to experiment with new tools and explore data sets. And public data repositories can link directly to these notebooks, making this a great way to interactively examine public data sets.

We deploy this infrastructure on Amazon Web Services, but using other environments should be easy as well. To deploy on AWS, we deploy a master service on a single instance and worker services on additional instances. The master runs the Apache web server and the load balancer (based on tmpnb-redirector, while the workers each run an instance of tmpnb.

git clone http://github.com/codeneuro/notebooks
cd notebooks

In the master directory, first run the setup scripts

cd master
make setup 

Deploy the Apache web server and start the screened redirector.py process:

make deploy

In the worker directory, first run the setup scripts:

cd worker
make setup 

Each worker must contain a conf.yml file which specifies:

• The hostname of the master node
• The port on the master node that handles registration commands (typically 9001)
• The hostname of the worker (so that the redirector redirects to readable URLs)
• The port of the worker's main tmpnb service (typically 8000)

A conf.yml template exists in the worker directory, and must be populated with the above values, specific to your deployment.

To start the worker, run

make launch 

If you want to build your own version, start with this repo and just modify the website content and docker images accordingly. We will work on simpler strategies for customization in the future.

This repo contains images for codeneuro/base and codeneuro/notebooks. The base image is based on jupyter/minimal, and the notebooks image sets up a custom scientific computing environment with local versions of tools like Spark and Thunder.