Failed to start matlabpool (under Cluster PBS using pbsdsh command)

Hi

I am using a "pbsdsh" command to run many similar MATLAB simulations. Each of the MATLAB program runs 8 parallel workers using 'local' scheduler (via SPMD()).

Each time I submit a job (that consists of the parallel jobs over 28 nodes, so total of 28x8=224 parallel runs), I am getting the following failures on some of the nodes. The occurrence and pattern of the error changes as I try multiple submission on the same fixed nodes. So for example, over the same nodes, sometimes 8 out of 28 nodes fails to start the matlabpool or sometimes more than 20 fails to start. Would you please take a look at the error message below and give me any tip? Or is there anyway I can start with fresh Matlabpool without seeing the errors/warnings?

example-1 ??? Error using ==> matlabpool at 133 Failed to start matlabpool. This is cause by: When setting up a parallel job, only 1 task is allowed To diagnose this issue further, validate the configuration 'local' in the Configuration Manager.

example-2 ??? Error using ==> matlabpool at 133 Failed to start matlabpool. This is cause by: A MatlabPool job must contain 1 task prior to submission To diagnose this issue further, validate the configuration 'local' in the Configuration Manager.

example-3 ??? Error using ==> matlabpool at 133 Failed to start matlabpool. This is cause by: The job was not submitted to the local scheduler because the MaximumNumberOfWorkers property was set to 0. The job would never start running with this value. To diagnose this issue further, validate the configuration 'local' in the Configuration Manager.

Sincerely, John FS

[Jamie]

John,

I've been having the same type of error messages (particularly example 3) for parallel jobs I've been submitting as well. Did you ever find a solution?

Thanks!

Jamie