In Rancher 2.8, upgrading the Kubernetes version on an RKE2 cluster originally provisioned in Rancher 2.7 triggers node rotation. We think this happens because Rancher 2.8 introduces the

gracefulShutdownTimeout

setting, which wasn't present in 2.7. As a result, Rancher detects a spec mismatch and replaces the nodes. We found this GitHub issue claiming that this rotation is required: https://github.com/rancher/terraform-provider-rancher2/issues/1222#issuecomment-1765466851 Is this still true? Is there any way to avoid the node rotation? We unfortunately have some applications that are pets and not cattle.