VPA Resource Right-Sizing Kubernetes

💡 Quick Answer: Deploy VPA in Off mode first to collect recommendations for 1 week without making changes. Review suggested CPU/memory values, then switch to Auto for non-critical workloads. For HPA-managed Deployments, use VPA for memory only (controlledResources: [memory]) to avoid conflicts.

The Problem

Teams guess at resource requests — usually 2-5x higher than needed, wasting 40-60% of cluster resources. Or they set them too low, causing OOMKills and throttling. VPA analyzes actual usage and recommends (or auto-applies) right-sized resource values.

The Solution

VPA in Recommendation-Only Mode

apiVersion: autoscaling.k8s.io/v1
kind: VerticalPodAutoscaler
metadata:
  name: api-server-vpa
  namespace: production
spec:
  targetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: api-server
  updatePolicy:
    updateMode: "Off"

After 1 week, check recommendations:

kubectl get vpa api-server-vpa -o jsonpath='{.status.recommendation}' | jq
# {
#   "containerRecommendations": [{
#     "containerName": "api",
#     "lowerBound": {"cpu": "50m", "memory": "128Mi"},
#     "target": {"cpu": "250m", "memory": "384Mi"},
#     "upperBound": {"cpu": "1", "memory": "1Gi"}
#   }]
# }

VPA Auto Mode

apiVersion: autoscaling.k8s.io/v1
kind: VerticalPodAutoscaler
metadata:
  name: worker-vpa
spec:
  targetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: background-worker
  updatePolicy:
    updateMode: "Auto"
  resourcePolicy:
    containerPolicies:
      - containerName: worker
        minAllowed:
          cpu: 100m
          memory: 128Mi
        maxAllowed:
          cpu: "4"
          memory: 4Gi

VPA + HPA Coexistence

# VPA: memory only (avoid CPU conflict with HPA)
apiVersion: autoscaling.k8s.io/v1
kind: VerticalPodAutoscaler
metadata:
  name: web-vpa
spec:
  targetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: web-server
  updatePolicy:
    updateMode: "Auto"
  resourcePolicy:
    containerPolicies:
      - containerName: web
        controlledResources: ["memory"]
---
# HPA: scale on CPU
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: web-hpa
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: web-server
  minReplicas: 2
  maxReplicas: 10
  metrics:
    - type: Resource
      resource:
        name: cpu
        target:
          type: Utilization
          averageUtilization: 70

graph TD
    VPA_OFF[VPA Off Mode<br/>1 week observation] -->|Review recommendations| DECIDE{Accept?}
    DECIDE -->|Yes| VPA_AUTO[VPA Auto Mode<br/>Live adjustment]
    DECIDE -->|Manual| APPLY[Apply recommended<br/>values manually]
    
    VPA_AUTO --> RESTART[Pod restarted<br/>with new resources]
    
    subgraph HPA + VPA
        HPA_N[HPA<br/>Scale replicas on CPU] 
        VPA_MEM[VPA<br/>Adjust memory only]
    end

Common Issues

VPA keeps restarting pods

Auto mode evicts pods to apply new resource values. Use updateMode: Initial to only set resources on pod creation (no restarts). Or use Off mode and apply manually.

VPA and HPA fighting over CPU

Never let VPA and HPA both control CPU. Use VPA for memory only: controlledResources: [memory].

Best Practices

• Start with Off mode — observe for 1 week before applying changes
• Set minAllowed and maxAllowed — prevent VPA from setting extreme values
• VPA + HPA coexistence — VPA for memory, HPA for CPU (never both on same resource)
• Auto mode for non-critical workloads — restarts pods to apply changes
• Initial mode for production — only applies on pod creation, no disruption

Key Takeaways

• VPA analyzes actual resource usage and recommends right-sized values
• Off mode collects recommendations safely; Auto mode applies them (with pod restarts)
• Most teams over-provision by 2-5x — VPA recovers 40-60% of wasted resources
• VPA + HPA coexistence: VPA controls memory, HPA controls horizontal scaling on CPU
• Always set minAllowed/maxAllowed bounds to prevent extreme recommendations