Unable to submit task result (Matlab parallel server)

Hi,

I am running some tests on a cluster. I create a job, and I submit several tasks. But, I get the following error

                Error: Cannot rerun task because there are no rerun attempts left (The task has no rerun attempts left.).
                       Original cancel message:
                       java.lang.Exception: Unable to submit task result - MATLAB will now exit and restart.
                       

Where shall I start to look at? What does practically this error mean? Is it a problem on the client side, or on the cluster side?

Answer 1

[Raymond Norris]

Hi Maria,

A few questions first:

• Which platform is MATLAB Parallel Server running on, Linux or Windows?
• Which scheduler are you using (MJS, PBS, etc.)?
• What size pool are you running?
• How many cores per node?
• How much RAM per node?

If you're running non-MJS, try the following. I'll show using both batch and parpool.

setenv('MDCE_DEBUG','true')
cluster = parcluster;

% If you're using batch
job = cluster.batch();
job.wait
cluster.getDebug(job)

% If you're using parpool
pctconfig('preservejobs',true);
pool = cluster.parpool();
cluster.getDebug(cluster.Jobs(end))

If you're using MJS

mjs = parcluster;
mjs.ClusterLogLevel = 4;
% Call either batch or parpool
mjs.getClusterLogs()

Perhaps the log file will display something else. If I had to guess, I'm betting you're running out of memory.