dandi / dandi-hub

Infrastructure and code for the dandihub
https://hub.dandiarchive.org
Other
11 stars 23 forks source link

Refactor Docker images to address a specific use case(s) and thereby reduce their size #143

Open kabilar opened 7 months ago

kabilar commented 7 months ago

Description

Currently, the DANDI Hub Docker images are quite large (~3.6 GB to ~13.8 GB) as they contain dependencies for several use cases. With the upcoming DANDI Hub refactor, users can select from a drop down menu of available images (see #124). So we can now create a series of images that are each tailored to a specific use case or a subset of use cases (e.g. for a given set of notebooks in the example-notebooks repository) and make them available in the drop down menu.

Requirements

  1. Define the majority of the use cases for the JupyterHub.
  2. Determine the dependencies common amongst subsets of use cases.
  3. Create Docker images for each subset of use cases defined in requirement 2.
  4. Image sizes should be minimized such that smaller user pods can be utilized.
  5. Add the images to the JupyterHub profileList.