Multiple Instances Of K8up

This guide describes how to install multiple instances of K8up on a cluster. The idea of having multiple instances is to enable upgrades between major versions over a longer period.

Limitations Of Multiple Instances

  • You can’t deploy multiple instances with the same major version of K8up.

  • It’s possible to encounter a name collision when using K8up’s PreBackupPod when both K8up v1 and v2 backups are configured for the same application.

Component Configuration

Sample configuration
  - backup-k8up as k8up-v1
  - backup-k8up as k8up-v2
    majorVersion: v1 (1)
    namespace: k8up-v1 (2)
    helmReleaseName: k8upv1 (3)
    majorVersion: v2 (1)
    namespace: k8up-v2 (2)
    helmReleaseName: k8upv2 (3)
1 Choose a different major version of K8up for each instance
2 Choose a different namespace for each instance
3 Choose a different Helm release name for each instance

You need to choose a different Helm release names since the component comes with cluster-scoped resources that share the same name. The ArgoCD apps will then "fight" against each other when doing syncs.

The backup-k8up.libjsonnet library provided by v3 of the component will render K8up resources for K8up v2. This causes Syn-managed K8up resources for K8up v1 to be deleted and replaced with K8up v2 by ArgoCD. In other words: Since objects are recreated you will lose all status and job information of past backups. However the actual Restic backup repositories should continue working.

Once both instances are installed kubectl get backup will return the Backup objects from in K8up v2. If you would like to list the Backup objects from K8up v1, run kubectl get (or for Schedules)


As outlined in more detail in Upgrade component from v2 to v3, deploying K8up v1 and v2 in parallel provides an upgrade window for users to migrate their backups from K8up v1 to v2.

Each application should be either working with K8up v1 or with K8up v2, but not both at the same time. For example, avoid upgrading Schedule to K8up v2 while leaving still having a PreBackupPod for K8up v1, otherwise the next scheduled backup will likely fail due to a missing PreBackupPod for K8up v2.

An upgrade for an application could look roughly as follows:

  1. Delete the old K8up v1 manifests from the cluster.

    Be sure that all child resources like Jobs and pre-backup Deployments are gone before continuing!
  2. Set the apiVersion for each K8up resources (for example Schedule) to in the source file (YAML/JSON spec)

  3. If applicable, update all K8up annotations used in your manifests (prefix changed from to

  4. Apply the new manifests

  5. Verify that backup still works