Using FSL with the Cluster

FSL <= 3.3

For FSL 3.3 or lower, clustered jobs are submitted using add on scripts written by our lab. If using the FSL GUI, ensure that you've clicked the

FEAT FMRI Analysis - CLUSTER 

button. When using FEAT from the command line, the following command is used:

$ feat_sge design.fsf

FSL >= 4.0

For FSL 4.0 or higher, SGE is integrated into the program itself by the FMRIB developers. As such, there is no special action that needs to be taken to ensure full usage of the cluster. As long as your using the BASH shell, your environment is fully ready to utilize clustered feat and all of the other lab utilities (freesurfer, fsl, UCLATools, etc).

A few "good practices"

Don't set programs as background processes (appending an ampersand)

It was common practice for many to set feat and the FSL GUI as a background process. This is done by adding an ampersand behind a command. For example,

$ fsl &

or

$ feat design.fsf &

While this seems convenient it also prevents meaningful information from being displayed in the terminal. FEAT or FSL could be throwing very pertinent error messages about flaws in the design file or analysis process and they would remain hidden. This frequently leads to confusion on why an analysis did not complete. A far superior alternative that provides the same conveniences as backgrounding a program and many, many more is the use of GNU Screen, a shell management program.

Always check your logs

Any time a analysis does not complete or the results don't look right always check your log output first. Though FSL provides a web based log, this is only a small part of the complete picture. Many times, problems not visible in these logs are revealed in others. For finding all your analysis logs when using clustered FEAT, refer to Finding FSL Debug Info.

Reading logs, or "debugging" in general is often the most tedious aspect of an otherwise joy filled endeavor. Unfortunately, it's a necessary evil not unlike Taxes, Death, and Human Resources.