There can be a number of causes for this, including, but not limited to the following:
- The cluster profile in MATLAB is configured to only use 1 core or 1 node
- The scheduler has only made 1 core or 1 node available to the person submitting the job
- There is only 1 usable core or node on the cluster
- You are not using MATLAB Parallel Server
First, review the settings in your cluster profile and make sure they are configured correctly. Settings and AdditionalProperties such as NumWorkers, NumThreads, ProcsPerNode, RequireExclusiveNode, Nodes, and NumCPUs can impact the number of cores and nodes being requested. Next, use scheduler commands (such as "squeue" or "scontrol show job <jobid>") and MATLAB Parallel Server log files to determine what information the scheduler is receiving for the job. To detect how many cores are usable on a node, run this command on the node:
Linux
matlab -dmlworker -r "feature numcores"
Windows
matlab -dmlworker -batch "feature numcores"
Please make sure that you are using MATLAB Parallel Server on your cluster. If you are manually writing a scheduler script that calls for MATLAB to be launched from the script itself, then this is not using MATLAB Parallel Server, just Parallel Computing Toolbox, and therefore, a job spanning nodes is unsupported. In order to use MATLAB Parallel Server, you must be submitting the job inside of MATLAB with MATLAB commands.
If everything else seems to be setup correctly, please try getting a new set of integration scripts from our website. When obtaining the scripts, download them through GitHub's GUI, rather than using "git clone".
Integration Scripts
