Skip to content

Upgrading a Cluster Installation

Find out Run:ai Cluster version

To find the Run:ai cluster version, run:

helm list -n runai -f runai-cluster

and record the chart version in the form of runai-cluster-<version-number>

Upgrade Run:ai cluster

Upgrade from version 2.9 or later

Run:

helm get values runai-cluster -n runai > old-values.yaml
  1. Review the file old-values.yaml and see if there are any changes performed during the last installation.
  2. Follow the instructions for Installing Run:ai to download a new values file.
  3. Merge the changes from Step 1 into the new values file.
  4. Run helm upgrade as per the instructions in the link above.

Note

To upgrade to a specific version of the Run:ai cluster, add --version <version-number> to the helm upgrade command. You can find the relevant version with helm search repo as described above.

Upgrade from version 2.8 or earlier

The process of upgrading from 2.7 or 2.8 requires uninstalling and then installing again. No data is lost during the process.

Note

The reason for this process is that Run:ai 2.9 cluster installation no longer installs pre-requisites. As such ownership of dependencies such as Prometheus will be undefined if a helm upgrade is run.

The process:

  • Delete the Run:ai cluster installation according to these instructions (do not delete the Run:ai cluster object from the user interface).
  • The following commands should be executed after running the helm uninstall command 
    kubectl -n runai delete all --all
kubectl -n runai delete cm --all
kubectl -n runai delete secret --all
kubectl -n runai delete roles --all
kubectl -n runai delete rolebindings --all
kubectl -n runai delete ingress --all
kubectl -n runai delete servicemonitors --all
kubectl -n runai delete podmonitors --all
kubectl delete validatingwebhookconfigurations.admissionregistration.k8s.io -l app=runai
kubectl delete mutatingwebhookconfigurations.admissionregistration.k8s.io -l app=runai
kubectl delete svc -n kube-system runai-cluster-kube-prometh-kubelet

  • Install the mandatory Run:ai prerequisites:

    • If you have previously installed the SaaS version of Run:ai version 2.7 or below, you will need to install both Ingress Controller and Prometheus.
    • If you have previously installed the SaaS version of Run:ai version 2.8 or any Self-hosted version of Run:ai, you will need to install Prometheus only.

  • Install Run:ai cluster as described here

Verify Successful Installation

See Verify your installation on how to verify a Run:ai cluster installation