 |
 |
> Good Afternoon, > > We've recently enabled hard memory limits on our gridengine queues to > prevent node oversubscription and subsequent death. This has been > working quite nicely, but it has also had the unintentional effect of > causing many fsl_sub-based jobs to die once they hit the default limit > (2G). > > Is there a way to specify complex values (e.g. 'qsub -l h_vmem=8G') > without needing to modify the fsl_sub script? Thank you very much for > your time. The fsl_sub script is intended to abstract away any local scheduling scheme so it should be the correct place to do this. Simply provide an FSL script h_vmem value here, then, if you intend to have different h_vmem for different script invocations, interrogate the passed-in command line to determine what value to use. This way you limit local changes to one script and won't have so many problems when you come to integrate the latest release/update of FSL. It's also more concise if you ever want to switch to a different scheduler. If you don't think this will work for you, e.g., you have to let users specify the value during their command line invocation, then you'd need to modify both fsl_sub and any calling scripts (NB. we can't offer therapy for anyone trying to modify the FEAT script). -- Dave Flitney, IT Manager Oxford Centre for Functional MRI of the Brain E:[log in to unmask] W:+44-1865-222713 F:+44-1865-222717 URL: http://www.fmrib.ox.ac.uk/~flitney