KCNA ยท Storage and troubleshooting
Troubleshooting nodes and the control plane
NotReady nodes, kubelet issues, API server and scheduler symptoms, system Pod logs.
Key points
- Start with kubectl get nodes to see which nodes are Ready.
- For a NotReady node, kubectl describe node shows conditions (Unknown with 'Kubelet stopped posting node status'), unreachable taints and a stale lease.
- Then check the kubelet and the network on that node.
- For control plane trouble, look at the logs of the API server, scheduler and controller manager.
Source: https://kubernetes.io/docs/tasks/debug/debug-cluster/
Practice questions
Which command gives a quick overview of whether the nodes are Ready?
- kubectl describe cluster
- kubectl get nodes
- kubectl logs nodes
- kubectl get pods --all
Show answer
Answer: B. kubectl get nodes
kubectl get nodes lists every node with its status. describe node is a deeper look at one node and kubectl logs reads container logs.
Which signals might appear when you describe a node that has stopped reporting?
- A PVC in the Lost phase
- A missing StorageClass
- node.kubernetes.io/unreachable taints
- Ready condition showing Unknown
- An old lease RenewTime
Show answer
Answer: C. node.kubernetes.io/unreachable taints, D. Ready condition showing Unknown, E. An old lease RenewTime
An unreachable node typically carries node.kubernetes.io/unreachable taints, conditions set to Unknown and a stale lease renew time. A Lost PVC and a missing StorageClass are storage matters, not node heartbeat signals.
True or false: as soon as a node becomes NotReady, all of its Pods are immediately evicted.
Show answer
Answer: False
False. Pods are evicted only after the node has been NotReady for a while (about five minutes in the docs' example), so they are still listed at first.
Practice 8 questions on this topic in Bitzu
Short lessons, instant explanations and spaced review of what you get wrong. Free on iOS and Android.
