Vertical Pod Autoscaler (VPA)

The Vertical Pod Autoscaler analyzes historical pod resource usage and produces CPU and memory right-sizing recommendations. It runs in recommend-only mode -- no pods are ever mutated automatically.

Details

Field	Value
Chart	vertical-pod-autoscaler
Repository	https://kubernetes.github.io/autoscaler
Version	0.8.1
Namespace	kube-system

Key Configuration

• Mode: Recommend-only. The updater and admission controller are disabled. Only the recommender component runs.
• Replicas: 1 (default is 2, reduced for homelab scale)
• Resources:
  • Requests: 25m CPU, 64Mi memory
  • Limits: 256Mi memory

How Recommendations Work

The recommender watches pods that have a corresponding VerticalPodAutoscaler CR and:

1. Collects CPU and memory usage from the Metrics API
2. Computes target, lower bound, and upper bound recommendations using percentile-based algorithms
3. Writes recommendations to the VPA object's .status.recommendation field
4. Exposes vpa_status_recommendation Prometheus metrics

Recommendations stabilize after several hours of usage data. Initial values should be treated as preliminary.

VPA CR Lifecycle

VPA CRs are not managed manually. Goldilocks automatically creates and manages a VPA CR (with updateMode: "Off") for every Deployment and StatefulSet in the cluster.

Applying Recommendations

VPA recommendations are advisory. To act on them:

1. Check the Goldilocks dashboard or query vpa_status_recommendation in Grafana
2. Compare the VPA target with the current request in the workload's values.yaml
3. Update the manifest and let ArgoCD sync the change

Upstream Documentation

https://github.com/kubernetes/autoscaler/tree/master/vertical-pod-autoscaler