Need to change location to look for kernels

JupyterLab
1

Hi,
I am working on a HPC cluster. We have two possible routes to get JL, JH, JN running:

  1. using our module system where modules are compiled against the cluster OS such that Anaconda, python etc and finally, say a python, R, or perhaps a Julia kernel is pip installed
  2. via a container, Singularity for instance, where the OS could be different from that in the cluster

Let’s say I load and run a JupyterLab container made from one of the images from the jupyter-stack. When I start in JL, I am presented with options for kernels. It seems these are loaded from:
~/.local/share/..../kernels/...

Is there some way to control this, some environment variable etc? If I choose a kernel that was compiled via the system OS, Centos, it usually fails.

Thanks,
Bob

2

@rsettlage one way to achieve could be to set the JUPYTER_PATH environment variable?

There is some documentation here: https://jupyter.readthedocs.io/en/latest/projects/jupyter-directories.html#data-files

3

Hi,
thanks. The JUPYTER_PATH variable allows me to set ADDITIONAL paths, but what if I want to LIMIT paths?

It seems to me that I want to do both. Set an additional path to push new kernels to and NOT display the default path so that incompatible kernels are not shown.

Thanks,
Bob