Hi Andi - Thanks for your kind words!
As explained on the trac-all wiki page (and also the help text you get from trac-all --help), the part about not submitting trac-all with pbsubmit is specific to our local Martinos Center users.
That's b/c we've modified fsl_sub (and bedpostx) to work on our local compute clusters (that run PBS). So trac-all recognizes when it's run on one of our clusters and submits each subject as a job. If it's run on any other machine, it won't do that, so this doesn't affect you. You're free to make your own modifications to fsl_sub or otherwise make use of your own cluster.
Let me know if you have any other questions, a.y
On Wed, 29 Feb 2012, Andi Heckel wrote:
Dear Tracula Team,
recently i ran trac-all (-prep / -bedp / -path), but I happened to submit the commands to our SGE computer cluster using fsl_sub (which is fsl's wrapper for qsub)
According to the header comments of the trac-all script, however, trac-all should not be submitted to a cluster. (it is self-submitting as i understand)
Do I need to worry about the results being invalid when submitting the command with qsub/fsl_sub or would you suggest re-running trac-all without fsl_sub / qsub, although all processes finished without errors according to the logs ?
Actually, i would really like to submit the trac-all command line using fsl_sub / qsub to a) make use of fsl_sub / qsub features and b) make full use of our cluster, so I can process multiple subjects in parallel, as it seems that not all time-consuming processing steps in the trac-all scripts are self-submitting (i guess only "trac-all -bedp" self-submits)
Are there general concerns, why one should not use fsl_sub/qsub on the trac-all (-prep / - path) command ? (I used fsl_sub on other self-submitting scripts, which seemed to work so far.)
Tracula is quite useful !
Thank you very much for your time ! Kind Regards, andi