This is a spack repository containing radio astronomy software, mostly for ASTRON's ARTS cluster.
Clone spack itself and this repository (n.b. this repo is probably not compatible with spack 1.0, so use the latest 0.x version), e.g.:
git clone -c feature.manyFiles=true --depth=2 https://github.com/spack/spack.git -b v0.23.1 /usr/local/spack
git clone https://github.com/TRASAL/spack /usr/local/spack/artsrepo
Enable spack in your current shell and bootstrap spack itself:
source /usr/local/spack/share/spack/setup-env.sh
spack bootstrap now
This repository comes with a spack environment file to do most of the setup automatically. To create an environment named "arts" from this file and activate it:
cd ~/spack-repo
spack env create arts env.yaml
spack env activate -p arts
The -p flag is optional and prepends the environment name to your shell prompt. A shorthand for spack env activate is spacktivate. Deactivation can be done with spack env deactivate or despacktivate.
Perform the final initialization steps inside the environment. These steps are
- Find already installed compilers on the system
- Find already installed packages on the system (e.g. CUDA)
- Ensure spack can find the packages defined in this repo
spack compiler find
spack external find --all -p /usr/local/cuda
spack repo add /usr/local/spack/artsrepo
N.b. is it possible to install other compilers using spack itself, but this takes a long time and the compilers on the ARTS system are new enough so these are used instead.
Now you are ready to install packages. To install all packages defined in the environment, first let spack resolve the dependencies:
spack concretize
If there are no errors, continue with the installation:
spack install
It is possible to run multiple instances of spack install in parallel to speed up the installation process.
There are two ways of using the spack environment: Through a spack view, giving access to all installed packages simultaneously, or through module files. Both methods are described here.
To use the spack view, first enable spack, followed by activating the arts environment:
source /usr/local/spack/share/spack/setup-env.sh
spack env activate -p arts
Now all installed packages are available in your current shell. If you want to know which packages are installed, run spack find. The relevant packages are listed under "root specs". Note: The environment includes Python 3.11. Several packages include python module that are also made available through the environment.
This method assumes Lmod is already installed on the system (it is on ARTS). Point Lmod to the spack modules to enable them:
module use /usr/local/spack/share/spack/lmod/linux-debian12-x86_64/Core
The exact path depends on the operating system and system architecture, the above is correct for ARTS.
Run module avail to get a list of available modules. In principle, a single version of each package is installed hence it is optional to specify the version number when loading a module. E.g. simply module load tempo2 works. Dependencies are also handle automatically, e.g. loading presto also loads tempo.
Note that several packages come with Python modules as well. When one of these packages is loaded, the Python module is loaded as well, but not the other way around. I.e. when you want to use e.g. the PRESTO Python modules, run either module load python followed by module load presto, or simply module load presto.
On the ARTS cluster the spack installation is made available locally on each node in /usr/local/spack. The installation is synced from arts041 to the other nodes. The installation path must be the same on all nodes to avoid breaking symlinks, wich are used internally by spack. An example script to sync the installation across the cluster is provided in bin/sync_spack.sh. It is advised to clean the spack cache with spack clean --all before syncing.