Setting up ipyparallel for using the linux clusters via ssh for jupyter

sshJupyter.md

Open a terminal that supports SSH (e.g. Git Bash). Check that login is successful:

ssh <user>@<ip of remote machine>

You should get prompted for passwword.

Password-less SSH

SSH alias

Create alias for ssh host by adding the following lines to <USERPROFILE>\.ssh\config (create this file without file extention if it doesn't exist)

Host <alias>
    HostName <ip of remote machine>
    Port 22
    User <user>

Now you can ssh using

ssh <alias>

authorisation

To set up for passwwordless login, first, generate a key for your local machine:

ssh-keygen -t dsa

The make a .ssh folder at the remote machine, if it doesn't exist:

ssh <alias>
mkdir .ssh
ls -a    # List all, check that ~/.ssh/ has been created.

Then copy the public key to the remote machine:

scp <USERPROFILE>\.ssh\id_rsa.pub <alias>:.ssh

SSH tunneling

Edit <USERPROFILE>\.ssh\config:

Host <alias>
    HostName <ip of remote machine>
    Port 22
    LocalForward 8000 127.0.0.1:8888    # 8888 is the default port for the first Jupyter notebook instance
    User <user>

Keep Jupyter Notebook running at remote after disconnection

Login via SSH, start the screen process, then Jupyter Notebook:

ssh <alias>
screen
cd <working dir>
jupyter notebook

You should now be able to access the Jupyter Notebook on the remote machine from your local browser at http://localhost:8000/.

Before logging out from SSH, save all unsaved input cells, then use the key combination Ctrl+A, quickly followed by D, to detach from the screen. After logging out, the tunnel http://localhost:8000/ is blocked, but will be recovered when you log in again. Any output generated while you are away will not be recoverable, but you can recall environmental outputs.