An evicted pod means Kubernetes killed it because the node ran out of resources — usually memory or disk space.

What causes this error

1. Node out of memory — pods on the node used more memory than available
2. Node out of disk — container logs, images, or volumes filled the disk
3. Pod exceeded its memory limit — OOMKilled then evicted

Fix 1: Check why it was evicted

kubectl describe pod my-pod | grep -A3 "Status"
# Look for:
# Status: Failed
# Reason: Evicted
# Message: "The node was low on resource: memory"

Fix 2: Set proper resource limits

resources:
  requests:
    memory: "256Mi"   # Guaranteed minimum
    cpu: "100m"
  limits:
    memory: "512Mi"   # Maximum before OOMKill
    cpu: "500m"

Without limits, a single pod can consume all node memory and trigger eviction of other pods.

Fix 3: Fix disk pressure

# Check node disk usage
kubectl describe node my-node | grep -A5 "Conditions"

# Clean up on the node
docker system prune -af  # Remove unused images
journalctl --vacuum-size=500M  # Trim logs

Fix 4: Add monitoring

Set up alerts for node resource usage before evictions happen:

• Memory usage > 80% → warning
• Disk usage > 85% → warning
• Pod restarts > 3 in 10 minutes → alert

How to prevent evictions

• Always set memory requests and limits on every pod
• Use PodDisruptionBudgets for critical workloads
• Monitor node resources with Prometheus/Grafana
• Set up cluster autoscaling to add nodes when resources are low

Related: Kubernetes kubectl cheat sheet · Kubernetes: Pod Stuck in Pending fix · Kubernetes: OOMKilled fix · What is Kubernetes