Skip to content

Upgrading

Upgrade the CRD chart first, then upgrade the operator chart.

helm upgrade --install omni-cluster-operator-crds \
  oci://ghcr.io/texas-hpc/charts/omni-cluster-operator-crds

helm upgrade --install omni-cluster-operator \
  oci://ghcr.io/texas-hpc/charts/omni-cluster-operator \
  --namespace omni-cluster-operator-system \
  --create-namespace

If you want a specific release, choose versions from the CRD chart package and operator chart package, then pass --version <chart-version> to both commands. When pinning versions, use matching versions for the two charts.

Before upgrading

  • Confirm cert-manager is installed and healthy.
  • Read the release notes for API or webhook validation changes.
  • Back up GitOps manifests for omni.texashpc.com resources.
  • Check that all existing resources are reconciled before changing the controller.
kubectl get omniconnections,omniclusters,omnicontrolplanes,omniworkers,omnimachines \
  --namespace omni-cluster-operator-system

Image overrides

For a branch build or unreleased test image, override the image tag:

helm upgrade --install omni-cluster-operator \
  oci://ghcr.io/texas-hpc/charts/omni-cluster-operator \
  --namespace omni-cluster-operator-system \
  --create-namespace \
  --set image.tag=dev

If the test image expects a specific chart release, add --version <chart-version>.

Rollback

If the operator deployment fails after an upgrade, roll back the operator chart first:

helm rollback omni-cluster-operator \
  --namespace omni-cluster-operator-system

Only roll back CRDs when you know the previous CRD version is compatible with the objects already stored in the Kubernetes API server.