You need to upgrade your Kubernetes control plane. What is the recommended approach for minimizing downtime?

Correct answer: Use a rolling upgrade strategy with backup copies

Option D is correct because Kubernetes recommends a rolling upgrade strategy where you maintain backup copies of control plane components and upgrade them one at a time, ensuring the cluster remains operational and recoverable at each step. This approach preserves high availability by keeping at least one instance of each component running throughout the process. Option A is wrong because upgrading all components simultaneously creates a window where none are available, causing total downtime and eliminating any rollback safety net. Option B is wrong because stopping all pods before upgrading introduces unnecessary downtime and is not how production Kubernetes upgrades are performed. Option C is wrong because the recommended Kubernetes upgrade order is to upgrade kubeadm first, then the control plane components (kube-apiserver, kube-controller-manager, kube-scheduler), then etcd, then kubelet and kubectl on each node; etcd is not the first component to upgrade.