TODO: Update the jupyterhub kernel JSON to point to the new location TODO: Add scripts for all the other packages TODO: Test that $CONDA_ENABLEMPI is actually working in the ./build_python_all script
Building Conda Python
- To rebuild the complete environment call: ./built_python_all.sh
- The install location is determined by the $CONDAPATH environment variable. If the variable is not set, then the default /global/common/software/m2650/python-$NERSC_HOST will be used. To customize the location where the install should be created, set the $CONDAPATH environment variable before calling the scripts.
- By default mpi and parallel HDF5 are disabled to avoid problems with running on NERSC login nodes. To enable parallel hdf5 and mpi set the environment variable $CONDA_ENABLEMPI before running the ./build_python_all script
- To install a particular package run the corresponding bash script from the packages folder
Installing a custom Kernel for Jupyterhub
The example Kernel for OpenMSI is located in the jupyterhub/ folder. To add the OpenMSI Jupyter kernel to jupyter.nersc.gov, simply copy jupyterhub/opemsi_nersc_jupyter_kernel.json to $HOME/.ipython/kernels/openmsi/kernel.json
NOTE: You have to create the kernels/openmsi folder first.