LAB 02 · EASY · 5–10 MIN

Pending Forever

The pod never schedules. There is capacity on the node, but the scheduler keeps saying no.

Goal: Schedule the workload onto a node tainted breakfix=special:NoSchedule.

Setup

kubectl get nodes
kubectl label node <NODE> breakfix-target=yes
kubectl taint node <NODE> breakfix=special:NoSchedule
kubectl apply -f broken.yaml

View broken.yaml · cleanup.sh

Hint 1

A Pending pod is a scheduler problem until proven otherwise. Read Events.

Hint 2

Capacity is not the only scheduler predicate. Inspect the target node's taints.

Solution

kubectl get pods -n breakfix-02
kubectl describe pod -n breakfix-02 -l app=worker
kubectl describe node <NODE> | sed -n '/Taints:/,/Unschedulable:/p'

The node has a NoSchedule taint and the pod has no matching toleration. Capacity is irrelevant until the scheduler is allowed to put the workload there.

Add a matching toleration to the pod template. Do not remove a production taint just to make one workload shut up.

tolerations:
- key: breakfix
  operator: Equal
  value: special
  effect: NoSchedule

Verify with kubectl get pods -n breakfix-02 -o wide.

Keep this: Pending is a scheduler symptom, not a generic Kubernetes mood. Read Events first.

Next lab: The PVC Never Binds →