Version v0.5 of the documentation is no longer actively maintained. The site that you are currently viewing is an archived snapshot. For up-to-date documentation, see the latest version.

Upgrading Kubeflow Deployments

Upgrading your Kubeflow deployment

Until version 1.0 Kubeflow makes no promises of backwards compatibility or upgradeability. Nonetheless, here are some instructions for updating your deployments.

Upgrading Kubeflow

Updating your deployment is a two step process:

  1. Update your ksonnet application:

    1. We recommend checking your app into source control to back it up before proceeding.
    2. Use the script upgrade_ks_app.py to update your ksonnet app with the current version for the Kubeflow packages.

    Note: ksonnet is working on support for this capability. See https://github.com/ksonnet/ksonnet/issues/237

  2. Update the actual deployment:

    1. Delete TFJobs v1alpha1 because Kubernetes can’t deploy multiple versions of a CRD:

      kubectl delete crd tfjobs.kubeflow.org
      
    2. Redeploy Kubeflow:

      ks apply ${ENVIRONMENT} -c ${COMPONENT}
      

Upgrading or reinstalling Kubeflow Pipelines

Read the Kubeflow Pipelines upgrade guide to make use of the following capabilities:

  • Upgrade: You can upgrade your Kubeflow Pipelines deployment to a later version without deleting and recreating the cluster.
  • Reinstall: You can delete a cluster and create a new cluster, specifying the storage to retrieve the original data in the new cluster.