[Neurodebian-users] Parallelization woes - condor and SGE

[Michael Hanke]

Hi,

[ let's take this off this list and move to neurodebian-users (CC'ed) ]

On Wed, Mar 14, 2012 at 08:35:54PM +0100, Cornelius Werner wrote:
> In my case, however, condor doesn't seem to accept randomise_parallel
> calls. Everything installed ok under Ubuntu 10.04.4 "Lucid Lynx"
> (fresh install - 11.04 and 12.04beta did not work, Linux Mint Debian
> Edition neither, surprisingly). When idle, condor_status correctly
> indicates eight queues, all of them unclaimed. However, after starting
> a randomise_parallel call, nothing really happens. condor_q /
> condor_status show an empty pipeline and three "held" jobs. And that's
> where I am stuck. Any ideas? Is this particular to randomise_parallel?
> I did not test other multi-threaded jobs, yet.

I wouldn't be surprised if there is a bug, this adaptor is fairly new. I
tested it with FSL's own regression test and a simple Nipype workflow it
works in those cases. If you can give me a command line that I can use
to reproduce the failure I'll track down the problem.

Also, could you elaborate what did work on lucid, but not on the others?

BTW: the 'condor_history' command allows for querying submitted (and
potentially failed) jobs. With the -l option you can ask for full
details, including the command that it tried to run.

Thanks and keep the bug reports coming,

Michael


-- 
Michael Hanke
http://mih.voxindeserto.de