Conda and Spack Installation

The recommended way to install Conda and Spack.

This guide goes over the recommended way to install Conda and Spack in ExCL. If you are already familiar with the Conda and Spack installation process, then these tools can be installed to their default locations. One recommendation is to store the environment.yml and spack.yaml files in your git repositories to make it easy to recreate the Conda and Spack environments required for that project. The remainder of this page goes over the installation in more detail.

Installing Conda

With recent changes to the Conda license, we are unable to use the default conda channel without a paid license. You are still able to use conda/miniconda with the conda-forge repository, but you must change it from using the default repository. See Transitioning from defaults | conda-forge | community-driven packaging for conda and Saying Goodbye to Anaconda?. Finding a replacement for Conda | by Robert McDermott | Medium for some additional information. The recommend approach is now to use venv, uv, or Pixi for managing python environments. These approaches work better and avoid the license issues. See also Python | ExCL User Docs for more information on how to get started with Python.

See the Conda Docs for the latest installation instructions. I install Miniconda instead of Anaconda since I do not require the 3GB of included packages that come with Anaconda and I will be installing my own packages anyways.

# Download the installer
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh

# Run the installer
bash Miniconda3-latest-Linux-x86_64.sh

Follow the prompts on the installer screens. Accept the license agreements. Specify /home/$USER/conda as the installation location. Choose if you want the installer to initialize Miniconda.

Improving Conda Environment Solver Performance

To improve the performance of the Conda environment solver, you can use the conda-libmamba-solver plugin which allows you to use libmamba, the same libsolv-powered solver used by mamba and micromamba, directly in conda.

The quick start guide is below.

conda update -n base conda
conda install -n base conda-libmamba-solver
conda config --set solver libmamba

See Anaconda | A Faster Solver for Conda: Libmamba and Getting started — conda-libmamba-solver for more information.

Installing Spack

# Install spack by checking out the right branch to /home
git clone https://github.com/spack/spack /home/$USER/spack
cd /home/$USER/spack
git checkout releases/latest # or release/v0.16

# Install a spack compiler to use as the default
spack install gcc@9.3.0 
spack compiler add $(spack location -i gcc@9.3.0)

# Add Spack to bashrc.
cat >> ~/.bashrc << 'EOL'
# Setup Spack
if [ -f "/home/$USER/spack/share/spack/setup-env.sh" ]; then
   source /home/$USER/spack/share/spack/setup-env.sh
fi
EOL