All I did (in the cluster where rancher is installed) was: 1. scale down rancher to 0 replicas in the old namespace "XYZ" 2. change the "meta.helm.sh/release-namespace" annotation from the old namespace "XYZ" to "cattle-system" in the PriorityClass "rancher-critical" and ClusterRoleBinding "rancher" resources 3. delete the ingress in namespace "XYZ" 4. install rancher via helm chart in namespace "cattle-system" 5. wait 10 to 30 minutes 6. check that everything works and the rancher-managed clusters are all connected to the new rancher installation in namespace "cattle-system" 7. delete the old rancher installation in the old namespace "XYZ" 8. Done.