Bootstrapping cachemachine

By default, cachemachine doesn’t do any prepulling and doesn’t provide a useful menu for Notebook Aspect spawning. As part of bootstrapping a new environment, you will want to configure it to prepull appropriate images.

For deployments on Google Kubernetes Engine, you will want to use Google Artifact Repository (GAR) as the source of images. See Google Cloud Artifact Registry (GAR) integration for basic information and instructions on how to configure workload identity.

For Telescope and Site deployments that need special images and image cycle configuration, start from the summit configuration. Consult with Telescope and Site to determine the correct recommended tag and cycle number.

For other deployments that use the normal Rubin Notebook Aspect images, a reasonable starting configuration for cachemachine is:

autostart:
  jupyter: |
    {
      "name": "jupyter",
      "labels": {},
      "repomen": [
        {
          "type": "RubinRepoMan",
          "registry_url": "registry.hub.docker.com",
          "repo": "lsstsqre/sciplat-lab",
          "recommended_tag": "recommended",
          "num_releases": 1,
          "num_weeklies": 2,
          "num_dailies": 3
        }
      ]
    }

This prepulls the latest release, the latest two weeklies, and the latest three dailies, as well as the image tagged recommended. However, also see Ensure the recommended image is pre-pulled for information on how to ensure cachemachine knows the correct tag and description for the recommended image.