Dear All, I’ve got a really weird FSL SGE problem that I hope someone can help me with…..
Background: we have a cluster running Ubuntu 12.04 LTS, that has the NeuroDebian packaged FSL installed. I wanted finer control than offered by the standard recommended SGE install for Ubuntu/Debian, so I manually configured veryshort.q short.q, long.q, verylong.q and bigmem.q. I then had a poke around, and took a look at fsl_sub, which doesn’t use these queues. So I took a copy of the fsl_sub distributed by FSL and installed it onto our cluster. So far, so good. To test the system I ran a basic Feat analysis, with B0 correction and non-linear registration to the standard - it worked perfectly (specified on the GUI and clicked GO) So to really push the hardware, and test the number of queues I had configured, I then set off 63 Feat jobs….. which have completed - sort of…….. Although registration to standard, and the non-linear options were “ON” in the preconfigured template design.fsf files that I ran using # feat design.fsf, when I come to look at the results it’s clear that none of the non-linear fnirts have run. There are no errors, no mention to fnirt in any of the log files, the system has a plenty of RAM (16 cores, 64 to 128GB), and all jobs apparently completed successfully. I’m at a loss as to know where to start with this, so any help would be much appreciated. Hope you can help. Jon -- Jonathan Brooks (PhD) - Senior Research Fellow CRiCBristol, University of Bristol, St Michael's Hill, Bristol, UK T: +44(0)117-342-1512 W: http://www.cricbristol.ac.uk _______________________________________________ Neurodebian-users mailing list [email protected] http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/neurodebian-users
