MATLAB DIstributed Computing Server

Hi,

I have a PBS TORQUE/MAUI cluster set up.

Here's the MATLAB I'm trying to run:

%Initialise cluster:
cluster = parallel.cluster.Generic()
set(cluster,'ClusterMatlabRoot','/gaia/software/src/matlab-2014a/build/')
set(cluster,'JobStorageLocation','/tmp/generic')
set(cluster,'NumWorkers',64)
set(cluster, 'OperatingSystem', 'unix');
set(cluster, 'IndependentSubmitFcn', {@independentSubmitFcn, 'gaia.ucd.ie', '/gaia/scratch'});
set(cluster, 'CommunicatingSubmitFcn', {@communicatingSubmitFcn, 'gaia.ucd.ie', '/gaia/scratch'});
set(cluster, 'GetJobStateFcn', @getJobStateFcn);
set(cluster, 'DeleteJobFcn', @deleteJobFcn);
set(cluster,'HasSharedFilesystem',false)

%create a job:
job = createJob(cluster)

%do some work:
createTask(job, @rand, 1, {3,3});
createTask(job, @rand, 1, {3,3});
createTask(job, @rand, 1, {3,3});
createTask(job, @rand, 1, {3,3});
createTask(job, @rand, 1, {3,3});

%submit job:
submit(job)

%Get results:
results = fetchOutputs(job);   %This fails!

Here is the error message:

Error using parallel.Job/fetchOutputs (line 841) Outputs can only be fetched if the job is in State 'finished'.

In MATLAB, the job is stuck permanently in a "pending" state:

job

job = 

   Job

      Properties: 

                     ID: 46
                   Type: independent
               Username: ehynes
                  State: queued
             SubmitTime: Wed Jun 11 18:34:33 IST 2014
              StartTime: 
       Running Duration: 0 days 0h 0m 0s

        AutoAttachFiles: true
    Auto Attached Files: List files
          AttachedFiles: {}
        AdditionalPaths: {}

      Associated Tasks: 

         Number Pending: 5
         Number Running: 0
        Number Finished: 0
      Task ID of Errors: []

When I do `qstat` on the cluster's head node command line, I see the job submitted and it goes from state "Q to R to E to C" (queued, running, exiting, completed).

[Edric Ellis]

Which integration scripts are you using? Jobs getting stuck in state 'pending' usually means that the workers on the cluster didn't have access to the JobStorageLocation. You might also want to check the JobStorageLocation to see if any log-files made it back there - they might have more information.