steps for installing jupyterhub

README.md

These are my notes while installing jupyterhub on an AWS EC2 instance running Ubuntu 16.04

Prerequisites

• install python3, pip3
  • sudo apt-get update && sudo apt-get install python3 python3-pip

• - ~~~`pip3 install pew`~~~
  

• - ~~~run all `pip install *` commands below from this new environment~~~
  

• using pew as striked out above would create an environment within the user's home directory with the installed dependencies
  • this is not suitable in this situation since we're looking for global packages for all users on the system
  • just make sure to use pip3 instead of pip in all below commands
• Install nodejs using nvm
• install jupyter's jupyterhub and notebook server
• test commands are installed: command -v nvm && nvm list && jupyterhub -h && configurable-http-proxy -h

- ~~~follow instructions in Digital Ocean's suggestion in [this SO answer](https://stackoverflow.com/a/29903645/4126114)~~~
  - ~~~because of [nvm issue#43](https://github.com/creationix/nvm/issues/43)~~~
- ~~~launch jupyterhub (`sudo pew in TEST_JUPTERHUB jupyterhub`)~~~


Otherwise, follow [Using sudo to run JupyterHub without root privileges](https://github.com/jupyterhub/jupyterhub/wiki/Using-sudo-to-run-JupyterHub-without-root-privileges) (read below notes first)
- all commands are run from whichever user is logged into the server to begin with
- wherever `$USER` is encountered, replace it with the username that you are testing as a jupyterhub user
- `useradd -G ...` is currently `useradd --ingroup ...` on ubuntu 16.04
- ~~~`sudo pip install sudospawner` can be without `sudo` since using `pew` to manage environments (check above)~~~
  - ~~~this goes with replacing `/usr/local/bin/sudospawner` with something like `/home/ubuntu/.local/share/virtualenvs/SSOC18/bin/sudospawner`~~~
  - for details on why this is striked out, check the note above about `pew`
- softlink the installed node and node package for the http proxy
```
sudo ln -s /home/ubuntu/.nvm/versions/node/v10.5.0/bin/configurable-http-proxy /usr/local/bin/
sudo ln -s /home/ubuntu/.nvm/versions/node/v10.5.0/bin/node /usr/local/bin/
```
- in this option's case, just launch jupyterhub with `launch.sh` below


Finally, some basic security testing, check `test_security.sh` below

launch.sh

#!/bin/sh
# after following steps from https://gist.github.com/shadiakiki1986/22bdddb4293b672145bae431164a1267
# and choosing the final step to be "Using sudo to run JupyterHub without root privileges"
cd /etc/jupyterhub
sudo -u rhea jupyterhub --JupyterHub.spawner_class=sudospawner.SudoSpawner

test_security.py

import subprocess

# system calls that do not need to fail
subprocess.call(["ls", "-l"])
subprocess.call("exit 1", shell=True)
subprocess.check_output(["echo", "Hello World!"])
subprocess.check_output(["whoami"])
subprocess.check_output(["ls", "-al", "/etc/jupyterhub/jupyterhub_cookie_secret"]) # The "cat ..." counterpart should fail. Check below

# system calls that should fail
subprocess.check_output(["sudo", "ls"])
subprocess.check_output(["cat", "/etc/sudoers"])
subprocess.check_output(["cat", "/etc/jupyterhub/jupyterhub_cookie_secret"])
subprocess.check_output(["touch", "/home/ubuntu/delete_me_1"])
subprocess.check_output(["rm", "/home/ubuntu/delete_me_2"]) # this is a file that was created by the user "ubuntu"