For this lab, we'll be running an IPython notebook in an interactive job on the Lion-XV cluster. Most of the instructions are there but we have to play a little bit of SSH magic to get this to work properly. In all of the steps below, change PORT to your unique assigned port number.

First, you need to SSH into the login node and create an SSH tunnel to forward the correct port back to your local machine. To do this, run the following command from your local terminal (don't forget to substitute the correct port number and your user name):

# On your local machine:
ssh -L PORT:localhost:PORT [email protected]

On the cluster, cd into your work directory and grab the code that you'll need from the GitHub repository:

# On the cluster login node:
cd work
git clone https://github.com/dfm/gp
cd gp

Then we'll start up an interactive job using PBS (asking for 2 hours just to be safe):

# On the cluster login node:
qsub -I -l nodes=1:ppn=1 -l walltime=2:00:00 -q WHAT_QUEUE_DO_WE_USE

Once that job starts up, load the correct Python module:

# In the interactive job:
module load python/2.7.3
cd $PBS_O_WORKDIR

Start a "reverse" SSH tunnel (just trust me on this one; remember to change the port number):

# In the interactive job:
ssh -f -N -R PORT:127.0.0.1:PORT [email protected]

And start up IPython (the port... remember):

ipython notebook --no-browser --matplotlib=inline --port=PORT

Finally, on your local machine, open up a web browser and point it at the URL: http://localhost:PORT (replacing PORT with the right number).