Parallel Job causing memory leak?

I have converted a script into a parralel job with a pretty simple outline. Roughly this:

---------------------

sched = findResource('scheduler', 'type', 'local'); set(sched,'ClusterSize',6);

for a = 1:nLoops

    job = createJob(sched);
    % create 6 tasks, start schedular with script, etc.
    % create sub-directories for each
    % collect results
    % kill job
end

-----------------

I have a 4 core machine with multi-threading and 12 GB of memory. Every time I run this script it eats more and more memory, until it crashes. Then I can't free up the memory unless I restart the computer (even leaving Matlab doesn't do it).

The script runs find outside the loop on a single core. Neve any issue that way. I see others have recently run into something similar. Is there a known issue with the Parallel computing toolbox? Am I doing something wrong?

Thanks!

Chris

Answer 1

[Jason Ross]

When you kill the job, are you using destroy?

[Chris DeVries]

Yes, I say:

destroy(job)

then the loop starts over at

job = createJob(sched)

The problem might be with a system command that I am calling in the script somewhere. I have to test that... but I think it's in the parallel computing toolbox.

Chris

[Jason Ross]

What's actually eating memory? If you look at Task Manager you'll see a few different MATLABs running. Do they continue to grow?

What version are you running?