Jupyter Lmod/Tmod

Jupyter interactive notebook server extension that allows user to interact with environment modules (Lmod or Tmod) before launching kernels. The extension use environment module's Python interface to accomplish module related task like loading, unloading, saving collection, etc.

requirements

• jupyter notebook >= 6.0
• Lmod >= 6.0 or Tmod >= 5.0
• optional: jupyterlab >= 3.0
• optional: jupyter-server-proxy >= 3.2.0
• optional: jupyterlab-server-proxy >= 3.2.0

Note that the extension supports Tmod < 5.0 too. However, if MODULES_RUN_QUARANTINE is not empty on the platform, module's Python API does not have correct behaviour. On default installations, MODULES_RUN_QUARANTINE=LD_LIBRARY_PATH is used. If LD_LIBRARY_PATH is not empty before loading a module, the existant paths in LD_LIBRARY_PATH is lost after loading the module. More discussion can be found here.

If jupyter-server-proxy and jupyterlab-server-proxy are detected, jupyter-lmod will add the proxy server launchers to JupyterLab UI when modules with matching names are loaded.

setup

install

pip install jupyterlmod

Disable jupyter-server-proxy notebook and lab extensions

To avoid having items in the launcher that cannot be launched because the binaries location are not in PATH, jupyter-lmod hides launcher items that do not have a corresponding loaded module. jupyter-server-proxy notebook and lab extension always display the launcher item. To avoid a situation where an item would be shown twice, we recommend disabling jupyter-server-proxy notebook and lab extensions.

This can be done with the following command for notebook:

jupyter nbextension disable --py jupyter_server_proxy --sys-prefix

and with the following commands for jupyterlab:

jupyter labextension disable @jupyterlab/server-proxy
jupyter labextension disable jupyterlab-server-proxy

Pinning launcher items

If server proxies do not have a corresponding modules, or you wish to have their launcher items displayed regardless of the loaded modules, you can define a list of items that will be pinned in the Jupyter notebook configuration file, like this:

c.Lmod.launcher_pins = ['Desktop', 'RStudio']

or

c.Tmod.launcher_pins = ['Desktop', 'RStudio']

based on your module system.

Mapping launcher items to modules

If the name of the server proxies do not match a corresponding module, or you wish to have multiple launchers for a single module, you can define a map of server proxies to module names in the Jupyter notebook configuration file, like this:

c.Lmod.launcher_module_map = {'RStudio': ['rstudio-server'], 'paraview': ['paraview-client', 'paraview/5.13']}

or

c.Tmod.launcher_module_map = {'RStudio': ['rstudio-server'], 'paraview': ['paraview-client', 'paraview/5.13']}

based on your module system.

JupyterHub and loading module that add kernels

If you have modules that modify JUPYTER_PATH and you access Jupyter through JupyterHub, make sure that c.Spawner.disable_user_config = False.

When disable_user_config = True, JupyterHub single-user server monkey-patches the jupyter_core.jupyter_path function to remove any path related to home. In order to do that, they retrieve the value of JUPYTER_PATH, remove any paths related to home, then keep it in a global variable. The jupyter_path function is then patched to only return versions of the global variable, making the function no longer affected by changer to JUPYTER_PATH.

The JupyterHub monkey-patching of jupyter_path can be read here.

demo

develop

requirements

• pip >= 23
• build
• nodejs >= 18.x

build

• wheel and tarball:

  pyproject-build
  

• labextension

  npm install
  npm run build
  # To install extension in jupyterlab in develop mode:
  npm run install:extension