Kubernetes Pod Stuck in Pending (Nothing Is Wrong With It)

Pending means the scheduler cannot find a home for the pod. Three usual suspects: no resources free, taints repelling it, or an unbindable volume.

What you'll see

Root causes

Insufficient CPU/memory on every node

Requests exceed what any node has allocatable. Event: 'Insufficient cpu'.

Taints without matching tolerations

Nodes carry taints (e.g. master/control-plane) that repel the pod. Event: 'node(s) had untolerated taint'.

PVC cannot bind

No StorageClass or no capacity. Event: 'volume node affinity conflict' or 'no persistent volumes available'.

Fix it

  1. Read the exact scheduling failure
    kubectl describe pod <pod> | tail -8
  2. Check real headroom on the nodes
    kubectl describe nodes | grep -A5 'Allocated resources'
  3. Lower requests or scale the pool
    kubectl scale deployment <name> --replicas=2  # or fix the request in the manifest
  4. For taints, add the toleration or schedule to a worker pool
    tolerations:
  - key: node-role.kubernetes.io/control-plane
    operator: Exists
    effect: NoSchedule

Field note

Requests are what the scheduler believes, not actual usage. A pod requesting 8 cores will sit Pending on nodes that have 4 allocatable — even if they are 90% idle.

Common questions

How do I find why a pod won't schedule?

kubectl describe pod <name> — the Events section lists the scheduling failures in order: insufficient CPU/memory, taints/tolerations, affinity, or PVC binding. The last message is the binding constraint.

My cluster has free nodes — why still pending?

Free in capacity terms isn't schedulable: taints, nodeSelectors, affinity rules, or topology constraints exclude the nodes. The describe output names which filter removed them.

Related fixes

Ship it right the first time

Kustomize base with probes, PDBs, and zero-downtime rollouts already wired.

Kubernetes Production Blueprints — $27 →

One-time. Yours to modify. Instant download from the NinjaOps template store.