Parallel Configuration Validation: times out, but job finishes; what to do?

I'm trying to get configured to a cluster at my school. The setup is:

• My client machine and the cluster do not share a filesystem
• The headnode and the worker nodes on the cluster share their file system
• My client runs r2011b and the cluster runs r2011a

Thus I've been using the `/usr/local/matlab/r2011b/toolbox/distcomp/examples/integration/sge/nonshared` example to get set up. It appears to be mostly working:

When I start the validation, it successfully completes the 'Find Resource' step, and moves onto the 'Distributed Job' step. While it's do this, I login to the cluster and look at the running jobs, and I see that it has submitted a job to the queue, and I can watch it move from 'pending' to 'running' to 'finished'. So the job gets done. But then on my client, it times out and fails.

How do I fix this so that my client matlab session sees that the job is getting done?

Answer 1

[Edric Ellis]

You must use the same version of MATLAB on both the client and the cluster. (Although that may not be the only problem - when jobs appear to complete on the cluster but the client cannot see that, it's often a permission problem on the Data Location of the cluster).