Uploaded image for project: 'Open Data Hub'
  1. Open Data Hub
  2. ODH-562

Alternate deployment method for some components

XMLWordPrintable

    • Icon: Story Story
    • Resolution: Unresolved
    • Icon: Minor Minor
    • None
    • None
    • odh-manifest
    • None

      ODH currently embraces Kubeflow's usage of Kustomize manifests to deploy components. This is great for many components: easy to use and adapt by creating overlays and custom manifests.

      However, for some more advanced scenarios (typically for production installations), this proves somewhat impractical compared to other methods like Helm or Ansible. For example, creating redundant components for HA with a variable number of instances as a parameter is hard as there is no "loop" function. Or you don't have access to helpers/scripts to retrieve data dynamicallyto customize your deployment.

      Tricks can be applied with initContainers or Jobs that embed scripts which are able to perform some actions, but that's closer to dirty hacks than real solutions.

      In my experience, real production installations ends up with heavily customized manifests that don't even really benefit from the ODH operator any more.

      This story is a placeholder for an open discussion on we could/should provide (documentation, scripts, recipes,...) to deploy and use ODH components in another manner.

            rhn-support-jnakfour Juana Nakfour (Inactive)
            gmoutier@redhat.com Guillaume Moutier
            Votes:
            0 Vote for this issue
            Watchers:
            1 Start watching this issue

              Created:
              Updated: