Users who wish to use their own versions of software in OnDemand must configure their environment properly so OnDemand will recognize their private modules.
This assumes you've installed Anaconda Python in your group's project directory. If you have an alternate installation location and/or version, make the appropriate changes in the instructions below.
1. Edit the .bashrc file to include the directory to your private modules
vi /user/<username>/.bashrc
Add the module use command to point to your modules files in either your home or project directory:
# User specific aliases and functions
module use /projects/academic/<group_name>/modulefiles
NOTE: not all project directories are academic. Yours may be under /projects/industry or /projects/rpci. You may also have your private modules in your home directory: /user/<username>/modulefiles
Save and exit the editor. Source the file so that the changes will be recognized:
source ~/.bashrc
2. Some software might require additional environment variables when used in OnDemand. If that is the case for your software, we recommend you create a separate module(s) for use with OnDemand. For example, Anaconda Python and Jupyter Notebook will require an additional setting. Copy your current python module to a new one, make sure to name your module 'jupyter' and ensure the following lines are included:
whatis([[Anaconda Python 3.x]])
setenv("PYTHONDIR","/projects/academic/<group_name>/anaconda3")
prepend_path{"PATH","/projects/academic/<group_name>/anaconda3/bin",delim=":",priority="0"}
prepend_path{"LD_LIBRARY_PATH","/projects/academic/<group_name>/anaconda3/lib",delim=":",priority="0"}
load("cuda/10.0")
NOTE: Your directory names and software versions may be different than this example
Please see the TouseCondaenvironmentinJupytercustomappTouseCondaenvironmentinJupytercustomappsection titled "To use Conda environment in Jupyter custom app" in the "Using Conda Environments on the CCR cluster" Guide for more details on this setup and a link to a video that demonstrates how to do this.
TouseCondaenvironmentinJupytercustomapp
3. [Optional]: If you're using Jupyter Notebook and wish to access your project directory, we recommend creating a symbolic link in your home directory to the project directory. Instructions can be found here