A limit was added for kube-controller-manager resources. Without taking limits into consideration when dynamically setting resource requests we can prevent the kube-controller-manager from getting deployed.

    4.20 and later

    Always

    1. Create a HostedCluster with the following annotation to size the KubeControllerManager:
resource-request-override.hypershift.openshift.io/kube-controller-manager.kube-controller-manager: memory=2560Mi
    2. Wait for the HostedCluster to provision
    

    Nodes never join the cluster because the kube controller manager fails to run.

    Nodes join the cluster

The kube controller manager fails to reconcile with the following error:
  - lastTransitionTime: "2025-09-09T22:05:40Z"
    message: 'Deployment.apps "kube-controller-manager" is invalid: spec.template.spec.containers[0].resources.requests:
      Invalid value: "2560Mi": must be less than or equal to memory limit of 1Gi'
    reason: ReconciliationError
    status: "False"
    type: RolloutComplete