Getting Started

Getting an account:

For a VPN Account, visit https://cfn.upenn.edu/computing/wiki/doku.php?id=accounts_and_vpn and look at the section "get a UPHS account with VPN access". Likely, you do not have a UPHS account yet, and you will have to e-mail Margaret ([email protected]) with the following form filled out: https://cfn.upenn.edu/CfN2014VPNinstructions.docx

What you need to download:

-Pulse Secure: http://www.uphs.upenn.edu/network/ --> Penn VPN Preferred (MAC) if connecting from Penn issued computer (i.e. lab computer)

-cx2go: https://cfn.upenn.edu/computing/wiki/doku.php?id=x2go_remote_desktop

-x11 quartz (to open GUI applications like Freeview) : https://www.xquartz.org

Completing tutorials:

Using the CFN Cluster requires a small amount of knowledge using Linux. There are multiple tutorials on Linux, many of which are included on the CFN wiki website (which is at the top of the page).

http://linuxsurvival.com gives a good tutorial on the basics of Linux and its commands.

Commands to know: http://linuxsurvival.com/command-list/

How to get to the shell (Mac): command+space and search "terminal".

Terminal is where you will type commands to get into the Cluster.

Once you have a cluster account, and you have a VPN account, you can now log onto the cluster!

Using the Cluster

Steps to log on to cluster:

Connect to the hospital VPN through BIG-IP FIX. Click this, then hover over UPHS, then click connect. You type in your VPN account information here. Note that it may be different than your cluster information. Make sure you are connected.

Open a terminal window. Once a terminal window is open, you can log onto chead with the command line ssh -Y [email protected]. Enter your password. You are now in the terminal and should see something like this!

Hooray! From here, you can use the different applications, open Freesurfer, etc., look at different files and go onto the shared data directory for the lab.

Using x2GoClient

Log onto x2goclient. To do this, find the downloaded file and open the application. https://cfn.upenn.edu/computing/wiki/doku.php?id=x2go_remote_desktop under "using the x2go client" will help with how to start a session and open a terminal within x2go.

Once a terminal window is open, you can log onto chead with the command line ssh -Y chead-local. This is much better for GUI-ed applications.

Freesurfer tutorials, tips, & helpful tools:

http://surfer.nmr.mgh.harvard.edu/fswiki/Tutorials

http://vitallongevity.utdallas.edu/cnl/wp-content/uploads/2015/10/FREESURFER_GUIDE_CNL_2015Oct.pdf

The default cluster version of Freesurfer is 5.3.0. Change the line below in your ~/.bash_profile from 5.3.0 to 6.0.0 in order to use version 6.0.

 export FREESURFER_HOME=$CFNAPPS/freesurfer/6.0.0

Using Nipype on the Cluster:

Set Up

Update 5.26.17 The cluster default installation of Python has been updated to 2.7.9. Old instructions for modifying your path to include different Python versions has been moved to Modifying Python Versions below, and shouldn't currently be necessary. However, if you want to use a different version of Python than the default, you may find them useful.

Install the Linux version of Miniconda into your /home directory on the cluster, and use conda config --add channels conda-forge; conda install nipype to install Nipype. I installed all Nipype strong recommendations through conda install as well. Miniconda will add a line to your .bash_rc file to put it on the path, but the cluster uses the .bash_profile file, so copy the line from .bash_rc into your .bash_profile file.

Install NiPy using easy_install-2.7 -d /home/user/miniconda2 nipy. Before installing Nipy, add the Miniconda directory (/home/user/miniconda2/) to $PYTHONPATH.

Some dependencies for running Nipype scripts still weren't found when running nipype.tests(). You may want to install the pandas, paramiko, dipy, nilearn, seaborn and mock packages through conda install.

Finally, I added the /home/user/miniconda2/bin directory to the beginning of the $PATH variable, so that the version of Python running in Miniconda, rather than the version on the cluster, is found when you run Python commands.

Usage

You may need to request more memory than the default allotment to run Nipype scripts. Use qsub -l h_vmem=6.5G, s_vmem=6.0G in order to do so, more info can be found on the cluster wiki page. I needed 6G's to run the Nipype preprocessing script.

Modifying Python Version

Modify your ~/.bash_profile $PYTHONPATH to point to /share/apps/python/2.7.9 (or whichever version you're interested in), and alias python to the correct version with alias python=/share/apps/python/Python-2.7.9 in the same file. Also add /share/apps/python/Python-2.7.9/bin to your $PATH variable. Other modifications to your bash_profile file may be necessary to get Python 2.7.9 running as the default version. You want which python to come up with the Python 2.7.9 *

*These instructions were tested 2.25.17 by Ursula, but may not have all steps included, or some steps may be unnecessary.