LUMI container wrapper¶

The LUMI container wrapper is a set of tools which wrap software installations inside an Apptainer/Singularity container to improve startup times, reduce I/O load, and lessen the number of files on large parallel file systems.

Additionally, the LUMI container wrapper will generate wrappers so that installed software can be used (almost) as if it was not containerized. Depending on tool selection and settings, either the whole host file system or a limited subset is visible during execution and installation. This means that it is possible to wrap installations using e.g. mpi4py while still relying on the host provided MPI installation.

The LUMI container wrapper is a general purpose installation wrapper that supports wrapping:

Existing installations on the filesystem: Mainly to reduce the I/O load and improve startup times, but may also be used to containerize existing installations that cannot be re-installed.
Existing Singularity/Apptainer containers: Mainly to hide the need for using the container runtime from the user.
Conda installations: Directly wrap a Conda installation based on a Conda environment file.
Pip installations: Directly wrap a pip installation based on a requirements.txt file.

The LUMI container wrapper is NOT generally recommended for managing Conda/pip installations

For some use cases, the LUMI container wrapper is an excellent tool for managing Conda/pip installations. For other use cases, there are better alternatives. See the installing Python packages guide for an overview the recommended ways to manage Python installations, including Conda/pip installations.

The LUMI container wrapper has some limitations

Please be aware of the limitations when using the LUMI container wrapper before wrapping your installations.

The LUMI container wrapper is experimental software

As the LUMI container wrapper is still under development, some of the more advanced features might change in exact usage and API.

Examples of using the LUMI container wrapper¶

The tools provided by the LUMI container wrapper are accessible by loading the lumi-container-wrapper module that is available in the LUMI and CrayEnv software stacks.

$ module load LUMI
$ module load lumi-container-wrapper

Once the module has been loaded, you can use one of the front-end tools conda-containerize, pip-containerize, wrap-container, and wrap-install.

Wrapping a basic Conda installation¶

In order to wrap a basic Conda installation, create an installation directory and run the conda-containerize tool

$ mkdir <install_dir>
$ conda-containerize new --prefix <install_dir> env.yml

where env.yml is a Conda environment file.

This file can be written manually, e.g:

channels:
  - conda-forge
dependencies:
  - python=3.8.8
  - scipy  - nglview

or generated from an existing environment

$ conda env export -n <target_env_name> > env.yaml

Windows and MacOS will need to add the --from-history flag to the export command

or, alternatively,

$ conda list -n <target_env_name> --explicit > env.txt

Using the --explicit option only works if the existing environment is on a Linux machine with x86 CPU architecture. Otherwise the result will not be transferable to LUMI.

After the installation is done, you simply need to add the bin directory <install_dir>/bin to your PATH.

$ export PATH="<install_dir>/bin:$PATH"

Then, you can call python or any other executables, Conda has installed, in the same way as if you had activated the environment.

If you also need to install some additional pip packages, you can do so by supplying the -r <req_file> argument e.g:

$ conda-containerize new -r req.txt --prefix <install_dir> env.yml

where req.txt is a standard pip requirements.txt file.

The tool also supports using Mamba for installing packages. Enable this feature by adding the --mamba flag, e.g. conda-containerize new --mamba ...

Make sure that you have read and understood the license terms for Miniconda as well as any used conda channels before using the command:

End-to-end example of wrapping a Conda installation¶

Using the previous env.yml

$ mkdir MyEnv
$ conda-containerize new --prefix MyEnv env.yml

After the installation finishes, we can add the installation directory to our PATH and use it like normal.

$ export PATH="$PWD/MyEnv/bin:$PATH"
$ python --version
3.8.8
$ python3
Python 3.8.8 | packaged by conda-forge | (default, Feb 20 2021, 16:22:27) 
[GCC 9.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import scipy
>>> import nglview
>>>

Modifying a wrapped Conda installation¶

As the wrapped installation resides in a container, it cannot be directly modified. Small python packages can be added outside the container, in the usual way, using pip, but then the python packages are sitting on the parallel file system which is not recommended for larger installations.

To actually modify the installation inside the wrapping container, you can use the update keyword together with the --post-install <file> option which specifies a bash script with commands to run to update the installation. The commands are executed with the Conda environment activated.

$ conda-containerize update <existing installation> --post-install <file>

where <file> could e.g. contain

conda  install -y numpy
conda  remove -y pyyaml
pip install requests

In this mode the whole host system is available including all software and modules.

Wrapping a plain pip installation¶

Sometimes you don't need a full-blown Conda environment, or you may prefer to manage your python installations using pip. For this case you can use the pip-containerize wrapper via

pip-containerize new --prefix <install_dir> req.txt

where req.txt is a standard pip requirements.txt file. The above notes and options for modifying a Conda installation apply to pip installations as well.

Note that the python version used by pip-containerize is the first python executable found on the PATH, so it's affected by loading modules.

Note

This python used to installed pip packages cannot itself be container-based as nesting of containers is not possible.

Additionally, you may use the --slim argument, which will use a pre-built minimal python container with a much newer version of python as a base. Without the --slim argument, the whole host system is available. However, by using the --slim argument, the system installations (i.e /usr, /lib64 ...) are no longer taken from the host, but are instead taken from the minimal python container.

Wrapping existing containers¶

The LUMI container wrapper also provides a tool to generate wrappers for existing Apptainer/Singularity containers, so that they can be used transparently without the need for prepending singularity exec ..., or modify scripts if switching between containerized versions of tools.

wrap-container -w </path/inside/container> <container> --prefix <install_dir>

where <container> can be a filepath or any URL accepted by singularity (e.g docker//:, oras//:, or a local .sif file), and -w needs to be an absolute path (or comma-separated list) inside the container. Wrappers will then be automatically created for the executables in the target directories or for the target path.

Additional examples¶

More examples may be found in the LUMI container wrapper GitHub repository examples.

How it works¶

A description of how LUMI container wrapper works may be found in the LUMI container wrapper GitHub repository README file. A short discussion of how LUMI container wrapper works from the user perspective may be found in this GitHub issue.