Singularity Registry HPC (shpc) can be installed from pypi, or from source.
In all cases, a module technology is required such as lmod (install intstructions)
or environment modules (install instructions).
Having module software installed means that the
module command should be on your path.
Once you are ready to install shpc along your module software, it’s recommended that you create a virtual environment, if you have not already
The recommended approach is to install from the repository directly, whether you use pip or another setup approach, and to install a known release. Here is how to clone the repository and do a local install.
# Install release 0.0.24 $ git clone -b 0.0.24 firstname.lastname@example.org:singularityhub/singularity-hpc $ cd singularity-hpc $ pip install -e .[all]
or (an example with python setuptools and installing from the main branch)
$ git clone email@example.com:singularityhub/singularity-hpc $ cd singularity-hpc $ python setup.py develop
if you install to a system python, meaning either of these commands:
$ python setup.py install $ pip install .
You will need to put the registry files elsewhere (update the
registry config argument to the path), as they will not be installed
alongside the package. The same is the case for modules - if you install to system
python it’s recommended to define
module_base as something else, unless you
can write to your install location. Installing locally ensures that you
can easily store your module files along with the install (the default until you
change it). Installation of singularity-hpc adds an executable, shpc to your path.
$ which shpc /opt/conda/bin/shpc
This executable should be accessible by an administrator, or anyone that you want to be able to manage containers. Your user base will be interacting with your containers via Lmod, so they do not need access to shpc. If you are a user creating your own folder of modules, you can add them to your module path.
Once it’s installed, you should be able to inspect the client!
$ shpc --help
You’ll next want to configure and create your registry, discussed next in Getting Started.
Generally, remember that your modules will be installed in the
modules folder, and container
recipes will come from the remote registry shpc-registry by default. If you don’t want your container images (sif files) installed alongside your module recipes,
then you can define
container_base to be somewhere else. You
can change these easily with
shpc config, as they are defined via these
variables in the config:
$ shpc config set module_base /<DIR> $ shpc config set container_base /<DIR>
Also importantly, if you are using environment modules (Tcl) and not LMOD, you need to tell shpc about this (as it defaults to LMOD):
$ shpc config set module_sys tcl
You can also easily (manually) update any settings in the
$ shpc config edit
Take a look at this file for other configuration settings, and see the Getting Started pages for next steps for setup and configuration, and interacting with your modules.
You must have your container technology of choice installed and on your $PATH to install container modules.
The module is available in pypi as singularity-hpc, and this is primarily to have a consistent means for release, and an interface to show the package. Since the registry files will not install and you would need to change the registry path and module base (making it hard to update from the git remote) we do not encourage you to install from pip unless you know exactly what you are doing.