VPA CPU Lab

Welcome — this lab covers a CPU-focused Vertical Pod Autoscaler (VPA) workflow. The high-level flow mirrors the memory-focused lab but targets CPU:

Deploy a sample application.
Monitor CPU utilization.
Apply a VPA that produces recommendations (no automatic updates).
Run a CPU load test and validate VPA recommendations.

A slide titled "Lab Overview" listing three numbered steps: "Deploy sample application," "Monitor application resource usage," and "Apply VPA configuration and capture recommendations." A stylized pink computer icon with a DNA-like symbol is shown on the left.

What you’ll deploy

A simple Flask application will act as the CPU workload for this lab.
Initially the app is idle and uses very little CPU (typically ~1m).
A VPA will be created to produce CPU recommendations only (no automatic updates). The VPA will enforce a minimum of 100m and a maximum of 1000m CPU for the container and will be configured to control CPU only.

VPA manifest (vpa-cpu.yml)

apiVersion: autoscaling.k8s.io/v1
kind: VerticalPodAutoscaler
metadata:
  name: flask-app
spec:
  targetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: flask-app-4
  updatePolicy:
    updateMode: "Off"  # Set to "Auto" for automatic updates
  resourcePolicy:
    containerPolicies:
      - containerName: '*'
        minAllowed:
          cpu: 100m
        maxAllowed:
          cpu: 1000m
        controlledResources: ["cpu"]

Quick explanation of the important fields

Field	Purpose	Example / notes
`targetRef`	Which Deployment the VPA should observe	`name: flask-app-4`
`updatePolicy.updateMode`	Whether the VPA should automatically change pod requests	`updateMode: "Off"` (recommendations only). Change to `"Auto"` to apply updates automatically.
`resourcePolicy.containerPolicies`	Per-container limits and controlled resources	`minAllowed: cpu: 100m`, `maxAllowed: cpu: 1000m`, `controlledResources: ["cpu"]`

Apply the VPA

kubectl apply -f vpa-cpu.yml
# Output:
verticalpodautoscaler.autoscaling.k8s.io/flask-app created

Inspecting recommendations

With the app idle, the VPA typically recommends something near the configured minimum (100m) because current observed CPU usage is very low (≈1m).
After generating CPU load against the Flask app, re-check the VPA recommendations. In this lab the recommendation rose from 100m up to about 126m, tracking the increased observed CPU usage while respecting the configured minAllowed and maxAllowed bounds.

Useful commands

Check pod CPU usage:
- kubectl top pods
Check VPA status and recommendations:
- kubectl describe vpa flask-app
- kubectl get vpa flask-app -o yaml

Millicore reminder: 1 CPU = 1000m. So 100m = 0.1 CPU and 126m ≈ 0.126 CPU.

Caution when using updateMode: "Auto": automatic updates may restart pods to change requests. Use Auto in production only after validating recommendations and testing rollout behavior.

Step-by-step walkthrough

Deploy the Flask test application (Deployment + Service).
Observe current CPU usage (e.g., kubectl top pods) — idle pods often show ~1m.
Create the VPA manifest (vpa-cpu.yml) and apply it:
- kubectl apply -f vpa-cpu.yml
Inspect VPA recommendations while the app is idle:
- kubectl describe vpa flask-app
- Recommendations will typically be near the configured minAllowed.
Run a short CPU load test against the Flask service (tool of your choice).
Re-inspect the VPA recommendations and confirm they increased (e.g., from 100m to ~126m), staying within 100m–1000m.

Notes and tips

The VPA only recommends values when updateMode: "Off". To have the VPA apply changes automatically, set updateMode to "Auto".
VPA recommendations are based on observed usage over time — brief spikes may not immediately alter recommendations.
Use kubectl describe vpa <name> to see summary recommendation information and any events.

Links and references

Manual Scaling

Horizontal Pod Autoscaler (HPA)

Vertical Pod Autoscaler (VPA)

Cluster Proportional Autoscaler (CPA)

Kubernetes Event-Driven Autoscaling (KEDA)

Introduction

Watch Video

Practice Lab