Skip to main content
Pod status: Failed / Reason: Evicted

Kubernetes Pod evicted by node pressure

The kubelet terminated a Pod to reclaim a constrained node resource such as memory, filesystem space or inodes.

kubelet eviction managerNode resourceshigh
Exact message
The node was low on resource

Also searched as: Evicted · DiskPressure · MemoryPressure · ephemeral-storage

What it means

Eviction protects node stability. The Pod object remains failed while a controller may create a replacement; diagnosis must identify the pressure signal and whether requests, limits, local storage or node capacity are misaligned.

Typical trigger conditions

  • Memory availability below eviction threshold
  • Node filesystem or image filesystem pressure
  • Inode exhaustion
  • Excessive emptyDir or container writable-layer usage
Evidence before intervention

First diagnostic checks

1

Read the eviction message

The failed Pod status and events identify the constrained resource and observed threshold.

kubectl describe pod <pod> -n <namespace>
2

Inspect node conditions

Check MemoryPressure, DiskPressure and PIDPressure plus recent node events.

kubectl describe node <node>
3

Compare workload declarations

Review requests, limits, ephemeral-storage declarations, emptyDir limits and PriorityClass.

Version and platform notes

Eviction signals and container-filesystem behaviour vary by Kubernetes and container-runtime configuration.

Engineering signal

Follow Kubernetes failures and fixes

One useful weekly email with new error references, tools, integration notes and production lessons. No daily noise.