Duke-UNC Brain Imaging and Analysis Center
BIAC Forums | Profile | Register | Active Topics | Members | Search | FAQ
Username:
Password:
Save Password   Forgot your Password?
 All Forums
 Support Forums
 Cluster Support
 new mounting infrastructure for cluster
 New Topic  Reply to Topic
 Printer Friendly
Author Previous Topic Topic Next Topic  

syam.gadde
BIAC Staff

USA
421 Posts

Posted - Sep 27 2011 :  12:34:14 PM  Show Profile  Reply with Quote
We are taking the opportunity to introduce a new auto-mounting infrastructure for the cluster, which is intended to be more reliable and more usable than the previous version.

Most of you should not notice a difference. In fact, the new infrastructure has been running on the cluster for several weeks now, and is basically a drop-in replacement. Anyone interested in the details can read on.

The old method of specifying "-v EXPERIMENT=Dummy.01" on the qsub command line, and specifying all paths using the $EXPERIMENT environment variable in the standard BIAC cluster template script will still work.

However, you may now also specify paths in your cluster scripts using the following path:
/mnt/BIAC/SERVER/SHARE...

So, for example:
/mnt/BIAC/munin.dhe.duke.edu/MyPI/Dummy.01

MyPI/Dummy.01 is, of course, just an example. You can browse the paths (for which you have permission) from the interactive node. These paths are usable on interactive and non-interactive nodes.

The first time you access one of these paths on a given node, it will automatically mount itself, and after the path has not been used for at least an hour, it will be automatically unmounted. Simultaneous access by multiple jobs is also supported.

This has the following benefits:

  • You no longer have to run lnexp to mount an experiment on any of the interactive cluster nodes, as long as you access the experiment through /mnt/BIAC .

  • Any paths created using this method are accessible to all nodes. For example, FEAT users will now be able to create design files on the interactive nodes (if you access the experiment data using the new path structure) and then use the resulting design.fsf file directly without rewriting the paths to use the value of the $EXPERIMENT variable in your job cluster scripts. Note: you may still need to rewrite your .fsf files as usual if you are re-using them as templates for multiple subjects.

  • You may now access multiple experiment directories simultaneously.

  • We have not yet enabled FSL's parallel processing by default, but this will enable you to do so after it has been tested.



We hope this will be a useful addition to the cluster. If you have any questions, please feel free to post here.

Updated: Now using /mnt/BIAC.

Edited by - syam.gadde on Oct 07 2011 09:40:36 AM

syam.gadde
BIAC Staff

USA
421 Posts

Posted - Oct 07 2011 :  09:38:14 AM  Show Profile  Reply with Quote
/mnt/BIAC now uses the new mounting infrastructure and /mnt/BIAC2 will be phased out eventually. Above post will be updated.
Go to Top of Page

dvsmith
Advanced Member

USA
218 Posts

Posted - Oct 08 2011 :  3:07:36 PM  Show Profile  Visit dvsmith's Homepage  Reply with Quote
Thanks, Syam! Is there any way we can help with the testing of fsl_sub (i.e., FSL's parallel processing)? Some of the scripts in ANTS have a similar feature that should work automatically when we select that option.

(from ${ANTSPATH}buildtemplateparallel.sh)
Optional arguments:

-c: Control for parallel computation (default 1) -- 0 == run serially, 1 == SGE qsub, 2 == use PEXEC (localhost)
Go to Top of Page

petty
BIAC Staff

USA
453 Posts

Posted - Oct 08 2011 :  3:29:19 PM  Show Profile  Reply with Quote
you can test fsl_sub by setting an environmental variable in your script: export FSLSUBMIT=1

Its default is zero, which means do not submit.

I've only tested FIRST segmentation, which worked fine. I have no idea if anything will hit the memory threshold as of yet.

As far as ANTS, i'd have to look at it to see if anything additional is needed for submissions.
Go to Top of Page

dvsmith
Advanced Member

USA
218 Posts

Posted - Oct 08 2011 :  4:43:41 PM  Show Profile  Visit dvsmith's Homepage  Reply with Quote
OK, cool. So do we actually need to call fsl_sub explicitly or will it do it automatically like feat and other programs (http://www.fmrib.ox.ac.uk/fsl/fsl/sgesub.html)?
Go to Top of Page

petty
BIAC Staff

USA
453 Posts

Posted - Oct 08 2011 :  7:06:40 PM  Show Profile  Reply with Quote
no, you would just run things normally ... then anything they've pre-programmed to submit would run by default.
Go to Top of Page
  Previous Topic Topic Next Topic  
 New Topic  Reply to Topic
 Printer Friendly
Jump To:
BIAC Forums © 2000-2010 Brain Imaging and Analysis Center Go To Top Of Page
This page was generated in 0.49 seconds. Snitz Forums 2000