Skip to content

node(s) didn't match PersistentVolume's node affinity

The scheduler ruled out nodes because the pod’s PersistentVolume can only be used from certain nodes (usually one zone), and none of those nodes can take the pod.

Releases before the volume binding plugin was reworded printed node(s) had volume node affinity conflict for the same condition. You’ll still see that wording on older clusters and in most search results.

  1. Where does the volume live? kubectl get pv <pv-name> -o jsonpath='{.spec.nodeAffinity}'. Look for a zone label such as topology.kubernetes.io/zone.
  2. Which nodes are in that zone, and why can’t they take the pod? kubectl get nodes -L topology.kubernetes.io/zone, then check those nodes for capacity, taints, or cordons.
Cause How it happens
Zone has no free capacity Node pool scaled down, or other pods filled it
Zone’s nodes are cordoned or gone Maintenance, an outage, or a node pool replaced in another zone
Disk created in the wrong zone StorageClass uses Immediate binding, so the disk was created before the pod’s placement was known
Local volumes The volume lives on one specific node that’s gone

Add capacity in the volume’s zone, or restore the data (from snapshot or backup) to a new disk in a zone that has capacity. Prevent it with volumeBindingMode: WaitForFirstConsumer and node pools in every zone you use.

Read next: Storage and CSI.