This is a shell script that spins up several popular data science-y server environments on one box. This script is tested and verified on Ubuntu 14.04 and 16.04.
This environment is built for a fresh install of Ubuntu. It will update/upgrade all base packages and install all needed dependencies. Software packages installed include:
- R
- RStudio Server
- Shiny-Server
- Anaconda Python: Anaconda inclues Jupyter Notebook (formerly IPython Notebook).
- Jupyterhub: A multi-user server for Jupyter Notebook.
Download or copy the shell script and run it on your Ubuntu box under your home user:
sh ./data_science_workbench.sh
WARNING: Don't execute as root, it needs your home directory for a couple things.
These scripts are optimized for a headless server but may also be used on an Ubuntu desktop machine. If instlalling in a desktop enviornment, the following GUI tools are recommended:
- Rstudio IDE
- Anaconda launcher: Install and launch Anaconda development tools.
You will be dumped into a Tmux shell since the Anaconda environment needs a new shell session to take affect.
Jupytherhub isn't running by default. It can be invoked by typing jupytherhub in the terminal.
I have a file in the R folder of this repo called "install_kernel.R" you can download and run it by:
https://raw.githubusercontent.com/keberwein/Data_Science_Workbench/master/R/install_kernel.R
Rscript install.kernel.R
Or, you can fire up R, install the needed repos and install the IRkernel yourself.
R
install.packages(c('rzmq','repr','IRkernel','IRdisplay'),
repos = c('http://irkernel.github.io/',getOption('repos')), type = 'source')
IRkernel::installspec(user = FALSE)
There are Jupyter kernels for serveral other programming languages as well. You can find those at the Jupyter kernels page.
After the script finishes, you'll have a few things running on their default ports.
- RStudio: On
localhost:8787(Username and password set in the script) - Shiny-Server: On
localhost:3838(No username or password) - Jupyterhub not started by default but you can fire it up with the command
jupyterhub. Onlocalhost:8000(Username and password are the same as those of the Ubuntu user that ran the script.)
A few default locations and files have been altered to allow universal access by all users on the system.
- R: The
Renvironfile has been altered to create a unified package library that is readable by all users and Shiny-Server. The file is located atusr/lib/R/etc/Renviron. - Anaconda: Normally stored in a user's home directory, Anaconda is installed in
/opt/anaconda3. - Anaconda: A path has been added to your user's
.bashrcfile to make Anaconda your default for Python and Pip. The file is located athome/YOUR_USERNAME. The added path is at the bottom of the file and isexport PATH="/opt/anaconda3/bin:$PATH". This line must be added manually for each new user on the system. The path can be added by going to the new user's home directory and running:echo 'export PATH="/opt/anaconda3/bin:$PATH"' >> ~/.bashrc
By default Jupyterhub use basic authentication, although SSL is available. To set up SSL, see the Jupytherhub documentation.
Shiny and RStudio Server are differant stories. Shiny open-souce edition ships with no authentication whatsoever, and RStudio Server only wiht basic auth. Both offer "Enterprise" editions, which offer SSL.
The simple solution would be to use an Apache2 reverse proxy to add SSL to both, but I have a feeling that may violate the terms of service. I'm too lazy to read the entire TOS, so I'll just recomend you don't do it. They are both licensed under AGPL v3 if anyone is interested.
Data Science Box is based on Ubuntu 12.04 and uses IPython Notebook instead of Jupyter.
Data Science Toolbox is available as either a Virtualbox image or on AWS. Again, it uses older versions of IPython Notebook but it looks as if it's under active development.