Parallel Computing - how to run jobs from clients on servers with the same version.

After submitting a batch, I try to retrieve the computed data using

j.fetchOutputs{:}

However, the system throws the following error:

Error using preCJSJob

This job was submitted with version 6.11 of the Parallel Computing Toolbox. However

the current MATLAB Distributed Computing Server is version 6.10. You must run jobs

from clients on servers with the same version.

How exactly do I run "jobs from clients on servers with the same version?"

Answer 1

[Edric Ellis]

Firstly, you need to ensure that the correct version of MATLAB Distributed Computing Server (MDCS) is installed on your cluster. Then, you need to ensure that the ClusterMatlabRoot property is set up correctly to launch workers from the location of the matching version MDCS installation. So, you might need to do something like this

c = parcluster('myClusterProfile');
% Work out where R2017b has been installed on your cluster nodes, and then:
c.ClusterMatlabRoot = '/opt/MATLAB/R2017b';
% If that works, save the changes for later:
c.saveProfile();

If you do not have an R2017b installation available on the cluster, then you'll need to use an R2017a client MATLAB & Parallel Computing Toolbox.