Skip to content

Tykky

Intro

Tykky is a set of tools which make software installations on HPC systems easier and more efficient using Apptainer containers.

Tykky use cases:

  • Conda installations, based on Conda environment.yml.
  • Pip installations, based on pip requirements.txt.
  • Container installations, based on existing Docker or Apptainer/Singularity images.

Tykky wraps 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, Tykky 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's possible to wrap installations using e.g mpi4py relying on the host-provided MPI installation.

This documentation covers a subset of the functionality and focuses on Conda and Python. Most advanced use-cases are not covered here yet.

Warning

As Tykky is still under development, some of the more advanced features might change with respect to exact usage and API.

Tykky module

To access Tykky tools:

1) Usually it is best to first unload all other modules:

module purge

2) Load the Tykky module:

module load tykky

Conda-based installation

First, make sure that you have read and understood the license terms for Miniconda and any used channels before using the command.

1) Create a Conda environment file env.yml:

An example of a suitable env.yml file would be:

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

Info

The channels field lists which channels the packages should be pulled from to this environment, whereas the dependencies field lists the actual Conda packages that will be installed into the environment. Note that Conda uses a channel priority for determining where to install packages from, i.e. it tries to first install packages from the first listed channel. If no package versions are specified, Conda always installs the latest versions.

2) Create a new directory <install_dir> for the installation. /projappl/<your_project>/... is recommended.

3) Create the installation:

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

4) Add the <install_dir>/bin directory to your $PATH:

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

5) Now you can call python and any other executables Conda has installed in the same way as if you had activated the environment.

Using Jupyter with a Tykky installation

To use a Tykky installation with Jupyter, include correct conda package in your Conda environment file: jupyter for Jupyter Notebooks or jupyterlab for Jupyter Lab. Also additional JupyterLab extensions can be installed, for example jupyterlab-git or dask-labextension.

The best way to use Jupyter in Puhti is with Puhti webinterface. See Jupyter application page for details how to use your own Tykky installation with Puhti web interface Jupyter.

Pip with Conda

To install some additional pip packages, add the -r <req_file> argument, e.g.:

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

Mamba

The tool also supports using Mamba for installing packages. Mamba often finds suitable packages much faster than Conda, so it is a good option when the required package list is long. Enable this feature by adding the --mamba flag.

conda-containerize new --mamba --prefix <install_dir> env.yml

End-to-end example

Create a new Conda-based installation using the previous env.yml file.

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

After the installation finishes, add the installation directory to your 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 Conda installation

Tykky installations reside in a container, so they can not be directly modified. Small Python packages can be added normally using pip, but then the Python packages will be sitting on the parallel file system, which is not recommended for any larger installations.

To actually modify the installation, we 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 nglview
pip install requests

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

Pip-based installations

Sometimes you don't need a full-blown Conda environment or you might prefer pip to manage Python installations. In this case we can use:

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

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

Note that the Python version used by pip-containerize is the first Python executable found in the path, so it's affected by loaded modules.

Important: This Python can not be itself container-based as nesting is not possible!

An additional --slim flag exists, which will instead use a pre-built minimal Python container with a much newer version of Python as a base. Without the --slim flag, the whole host system is available, whereas with the flag the system installations (i.e. /usr, /lib64, ...) are no longer taken from the host, but instead coming from within the container.

Container-based installations

Tykky also provides an option to:

  • Generate wrappers for tools in existing Apptainer/Singularity containers so that they can be used transparently (no need to prepend apptainer exec ... or modify scripts if switching between containerized versions and "normal" installations).
  • Install tools available in Docker images, including generating wrappers.
wrap-container -w /path/inside/container <container> --prefix <install_dir>
  • <container> can be a local filepath or any URL accepted by Apptainer/Singularity (e.g docker:// oras://)
  • -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 / for the target path. If you do not know the path of the executables in the container, open a shell inside the container and use the which command. To open a shell:
    • In case of existing local Apptainer/Singularity file: singularity shell image.sif.
    • In case of Docker or non-local Apptainer/Singularity file, create first the installation with some path and then start with created _debug_shell.

Memory errors

With very large installations the resources available on the login node might not be enough, resulting in Tykky failing with a MemoryError. In this case, the installation needs to be done on a compute node, for example using an interactive session:

# Start interactive session, here with 12 GB memory and 15 GB local disk (increase if needed)
sinteractive --account <project> --time 1:00:00 --mem 12000 --tmp 15

# Load Tykky
module purge
module load tykky

# Run the Tykky commands as described above, e.g.
conda-containerize new --prefix <install_dir> env.yml

Moving and deleting Tykky installation

For deleting a Tykky installation, remove the folder.

Tykky installations can also be moved:

  • Inside the same supercomputer, from folder to folder, move the folder with mv to new location.
  • Between Puhti and Mahti use rsync. For copying to Mahti, log in to Mahti and change to the folder where you want to move the Tykky installation, then use: 
rsync -al <username>@puhti.csc.fi:<install_dir> .

More complicated example

Example in tool repository.

How it works

See the README in the source code repository. The source code can be found in the GitHub repository.