Running on HPC clusters (Torque/Slurm)

[namheegordonkim]

I'm using Torque and Slurm in conjunction with Jupyter. I managed to confirm these facts:

• When Jupyter is running on a head node (that is, not one of the HPC nodes but the login node connected to the rest of the Internet), matlab_kernel works well
• When Jupyter is running on an HPC node via interactive session, matlab_kernel does not work. I see Assertion failed: nbytes == sizeof dummyon Torque, and the kernel never successfully connects on Slurm (i.e. Replacing stale connection while doing something as simple as disp('hello'). I also observe matlab.engine.EngineError: Transport stopped.).
• When Jupyter is running on an HPC node via interactive session, Python and R kernels work well as intended.

I'm trying to figure out if this is an issue with a particular HPC configuration or an issue with matlab_kernel. Has anyone else experienced the same?