Farm customization

Or how to install a program, make it a module, create a virtual environment, and use them together in a script. Using miniconda3 and SMC++ as an example. (SMC++ has disappeared from farm's miniconda3 -AKA conda3/bio3- install more than once and a reliable installation was needed.) Disclaimer: Works as intended but may not be the best method

INSTALL THE PROGRAM

First, get the Miniconda3 install file

wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh

Install to your chosen directory

bash ./Miniconda3-latest-Linux-x86_64.sh -b -p $HOME/Software/miniconda3

• (flags: -b forced unattended installation; -p indicates the installation path)
• Note: Also see the more comprehensive information about software installation on farm

CREATE A PERSONAL MODULE

Create your personal modules directory

mkdir -p $HOME/MyModules/miniconda3

• Note: miniconda3 was not a module listed in module avail for the cluster therefore it was okay to use, however, I think you can name the module whatever you want.

Copy the module file for miniconda (i.e. conda3) to your personal modules directory

cp /share/apps/modulefiles/hpc/conda3/1.0 $HOME/MyModules/miniconda3

Modify the setenv in $HOME/MyModules/miniconda3/1.0 to the user installation

• Original: setenv CONDA3_HOME /share/apps/conda3/miniconda3

• Modified: setenv MINICONDA3_HOME /<full path to program here>/miniconda3

• Note: modified CONDA3_HOME throughout 1.0 file to MINICONDA3_HOME to ensure proper functioning

Add to your available modules

module use $HOME/MyModules

• Note: module avail will now show your personal modules under the first section listed, in my case /home/dmvelasc/MyModules
• Note: module use $HOME/MyModules and export MODULEPATH=$HOME/MyModules:$MODULEPATH only temporarily save $HOME/MyModules in the $MODULEPATH.

CREATE A VIRTUAL ENVIRONMENT

Load your miniconda module

module load miniconda3

• Note: Be sure to add your module to the module list with module use $HOME/MyModules before calling it

Create a virtual environments directory

mkdir $HOME/.virtualenvs

Create the desired virtual environment

virtualenv -p python3 .virtualenvs/smcpp

• Note: the -p flag specifies the python version to use, can be even more version-specific if that version is installed

Activate the virtual environment

source $HOME/.virtualenvs/smcpp/bin/activate

Install SMC++

conda install -c terhorst -c bioconda smcpp

• Note: DO NOT DO THIS ON THE HEAD NODE use an interactive session

USE YOUR NEW MODULE AND VIRTUAL ENVIRONMENT IN A SCRIPT

You will need to include the following lines, with full paths, in your script after the header, etc.

module use /<full path>/MyModules
module load miniconda3
set +u
source <full path>/.virtualenvs/smcpp/bin/activate
set -u

• Note: the set +u and set -u are workarounds to an unbound variable problem in the activate script when using set -u in the bash script

Then carry on as normal in your script with bash and SMC++ commands

Helpful resources used to put this together

• University of Utah Center for High Performance Computing User Installed Python and Advanced Modules Tutorial
• The University of Chicago Research Computing Center Tips and Tricks
• freeCodeCamp Why you need Python environments and how to manage them with Conda
• HiPer Gator Research Computing Personal Python Module Installation and Modules
• Torrant's Tech Blog Using Pythong Virtual Environments with Slurm