Migrate EKF Cluster

This guide will walk you through migrating all EKF resources of a source to a destination Arrikto EKF cluster. The EKF resources we support for migration are:

  • Machine Learning metadata (MLMD)
  • Notebooks
  • Pipelines
  • Models
  • Katib experiments
  • Kubeflow profiles
  • PVCs backed by the Rok storage class

First you will snapshot the EKF resources and publish them into a Rok Registry using our backup tool. Then, you will download the EKF resources and publish them to the destination EKF cluster using our restore tool. We provide two ways to run these tools:

  1. Interactive run: This means that during the process you will be prompted for input.
  2. Non-interactive run: This means that you will run the tools without being prompted for input.

Note

Arrikto supports the migration of a source EKF cluster of verion 1.3 or later to a destination EKF cluster of version 1.5.