study.

Investigating a NotReady Node · CKA

Start with node status, then narrow the investigation to the kubelet and runtime.

← CKA

Learn with progress and quizzes

First, understand the concepts.

Node conditions

Inspect conditions such as Ready, MemoryPressure, DiskPressure, and PIDPressure, along with Events. NotReady alone does not identify a specific cause.

Host investigation

Check kubelet logs, the container runtime, disks, and networking. Run host commands only when you have access to that node.

Why it matters

First distinguish one failing Pod from many failures on the same node. Scope helps separate application settings from a shared node problem.

Work through the reasoning

Read conditions and timing

Inspect Ready, MemoryPressure, DiskPressure, PIDPressure, transitions, and events. Unknown readiness can indicate a reporting-path failure; it does not by itself prove the machine is powered off.

Collect host evidence

On an accessible practice node, inspect kubelet and runtime state, recent logs, and filesystem capacity. Free bytes do not rule out inode exhaustion, so consider df -i too. Record observation times.

Learn by doing.

You need kubectl and access to your own practice Kubernetes cluster. Each lab is independent. Check your current context first, then run the commands in your own terminal.

Replace NODE_NAME with an actual node name. Run the host commands below only in a terminal connected directly to that systemd-based node.

kubectl get nodes
kubectl describe node NODE_NAME
# Run the following commands in a terminal on that node
sudo systemctl status kubelet --no-pager
sudo journalctl -u kubelet --since "10 minutes ago" --no-pager
df -h

How to check your result

Read the Ready condition and recent events. If you cannot access a managed node, skip host commands and check the provider’s node logging facilities.

Apply what you learned

Multiple Pods fail on one node with DiskPressure=True. What should you collect and avoid?

Reveal the reasoning after trying

Collect filesystem and inode usage, kubelet and runtime logs, affected Pods, and event timing. Do not manually delete runtime data directories. Choose supported cleanup and retention procedures after identifying the cause.

Ready to check your understanding?

What is the best first step when a node is NotReady?

  1. Immediately delete all Pods
  2. Check Node conditions and Events
  3. Delete every Service
  4. Replace certificates without investigating
Reveal the reasoning after trying

Check Node conditions and Events

Collect evidence from conditions and events first. Then narrow the investigation to the kubelet, runtime, networking, and resources.

Kubernetes documentation