e7dal / alpine-dask-docker

Tiny Dask Docker images based on Alpine Linux

Alpine Dask Docker Images

Here we provide a few Docker images for running Dask on Alpine Linux. These are designed to be as small as possible while still providing a good user experience.

Three images are provided:

• jcrist/alpine-conda

  A base image with miniconda installed.

• jcrist/alpine-dask

  An image for running both the dask schedulers and workers.

• jcrist/alpine-dask-notebook

  An image with all of the above, as well as Jupyter Notebook and JupyterLab installed.

These images can be used as is, or extended with additional packages by using them as a base image:

FROM jcrist/alpine-dask:1.2.2

# Add scikit-learn and numba, and cleanup afterwards
RUN /opt/conda/bin/conda install --freeze-installed -y \
        scikit-learn \
        numba \
    && /opt/conda/bin/conda clean -afy \
    && find /opt/conda/ -follow -type f -name '*.a' -delete \
    && find /opt/conda/ -follow -type f -name '*.pyc' -delete

Alternatively, additional packages can be installed at runtime via the EXTRA_CONDA_PACKAGES/EXTRA_PIP_PACKAGES environment variables:

$ docker run -e "EXTRA_CONDA_PACKAGES=scikit-learn numba" jcrist/alpine-dask dask-scheduler

These images should be drop-in usable with the existing Dask Helm Chart, with the benefit of being much smaller images.