Spack extension for automatic package configuration and deployment.
You can try out this Spack extension be executing 4 easy steps:
- Set up and activate a local python environment
- Set up and activate
spack - Install
spack-sdploydependencies - Clone and configure spack-sdploy
This 4 steps are now detailed in the next section.
Just for a matter of completeness, all the steps needed get up and running with spack-sdploy extension will be covered, which can be a bit pedantic.
It is recommended that a Python environment be used to support sdploy. This same Python can also be used to run Spack.
python3 -m venv <path-to-environment-directory>
. <path-to-environment-directory>/bin/activate
For more information on how to create a virtual environment in Python refer to the PEP 405 – Python Virtual Environments documentation.
See the Spack documentation on how to install Spack. For sake of completeness, we copy paste the commands here:
git clone -c feature.manyFiles=true https://github.com/spack/spack.git
. spack/share/spack/setup-env.sh
Up to now the only dependency of spack-sdploy if jinja2. Once you have activated Python environment, you can simply use pip to install the packages.
pip install jinja2
git clone [email protected]:epfl-scitas/spack-sdploy
To activate the spack-sdploy extension you must add it to the config.yaml. If you already have another Spack installation and just want to try out spack-sdploy may very well create a temporary directory to store the configuration and then use the SPACK_USER_CONFIG_PATH variable to point this new directory.
mkdir temporary_config
export SPACK_USER_CONFIG_PATH=/path/to/temporary_config
and then, inside the temporary_config directory, write a config.yaml file with the following contents:
config:
extensions:
- /path/to/spack-sdploy
Be sure you do not change the spack-dploy directory. Spack forces the extensions
to follow strict rules. Please see the
Spack Extensions
documentation for more details about this subject. At this point you should now
be able to call spack -h and see the new Spack commands deployed by the
spack-sdploy extension.
At the present time, spack-sdploy will add 2 commands to your already existing Spack commands. These commandes are:
spack write-spack-yaml
spack write-packages-yaml
In the future we may change the names of these commands, but for now lets just imagine these are short and easy to type commands.
As you may have guessed it (if you haven't that's ok), write-spack-yaml will write the spack.yaml file and write-packages-yaml will write the packages.yaml file. Of course, Spack does not (yet!) guess what you may want to install and for that purpose, both these commands will read all the specs you want in your spack.yaml file by reading another file you have previously written and which we call by stack.yaml.
For the time being, spack-sdploy already comes with a dummy stack.yaml so we can get started using the new commands.
spack write-spack-yaml
spack write-packages-yaml
spack write-activate-list -p <platform> -s <stack>
Write to file named packages_to_activate list of packages to activate, using spack activate <package>. Packages are writen one per line.
Packages to activate can be marked in the stack file in two possible ways: by adding the keyword activate: true in the metadata section of a list of packages or by adding the keyword activate: true to an individual package. Duplicates are removed.