Learn to upgrade a kubeadm cluster one minor version at a time, drain nodes safely, and back up and restore etcd with snapshots for the CKA exam.
It is the last Saturday of the quarter. Priya, a platform engineer in Pune, has a two-hour window to move her company's 12-node cluster from one...
Most failed upgrades were doomed before the first command ran.
The control plane upgrade runs in a fixed order: upgrade the kubeadm tool, let kubeadm upgrade the cluster components, then upgrade the node's own...
Workers are where outages actually happen, because they run your applications.
If the control plane is the brain of a cluster, etcd is its memory.
Restore is the rare, high-stakes operation.
It is the last Saturday of the quarter. Priya, a platform engineer in Pune, has a two-hour window to move her company's...
Most failed upgrades were doomed before the first command ran.
The control plane upgrade runs in a fixed order: upgrade the kubeadm tool, let kubeadm upgrade the cluster components...
Workers are where outages actually happen, because they run your applications.
If the control plane is the brain of a cluster, etcd is its memory.
Restore is the rare, high-stakes operation.
Kubernetes certificates created by kubeadm last one year.
Upgrade and restore failures follow a small number of patterns.
The CKA is a speed exam. These six drills mirror exam tasks.
This lab uses the multi-node kubeadm cluster you built in the Cluster Architecture module: one control plane node (cp1)...
Command quick reference The upgrade order in one line: back up, kubeadm, plan, apply, drain, kubelet, uncordon, then...
Roles that use the skills in this module.
DevOps Engineer
Site Reliability Engineer
Platform Engineer
Not a software engineer sheet. Every problem comes from real DevOps, SRE, Platform and Cloud interviews, from your first script to a system you build yourself.
Open the Coding SheetNo. kubeadm supports upgrading one minor version at a time, so 1.33 goes to 1.34 and then to 1.35. Patch versions within a minor can be skipped freely. If you are two minors behind, you run the full upgrade procedure twice.
Yes, drain it after kubeadm upgrade apply and before you upgrade its kubelet and kubectl packages. Draining moves regular workloads off the node while the kubelet restarts. Uncordon it once the new kubelet reports Ready.
The usual causes are DaemonSet pods, pods using emptyDir volumes, unmanaged pods with no controller, and PodDisruptionBudgets that would be violated. Each one has a flag or a fix, and you should read the error message before reaching for --force. A PodDisruptionBudget block usually means you should scale the app up or wait.
A snapshot captures all API objects stored in etcd, such as Deployments, Secrets and ConfigMaps. It does not capture the data inside your PersistentVolumes, nor node-level files such as certificates and kubeadm configuration. For real recovery you back those up separately.
Current etcd releases use etcdutl for offline snapshot work, including snapshot restore and snapshot status. The etcdctl snapshot save command still talks to a running server and is the right tool for taking the backup. Older etcdctl restore examples still circulate online, so check which version your cluster runs.