Skip to content

VPA update modes

VPA update modes determine how the Vertical Pod Autoscaler (VPA) applies resource recommendations to Pods. The mode is configured using the spec.updatePolicy.updateMode field within a VPA resource definition.^[400-devops__06-Kubernetes__k8s-ithelp__Day27__README.md]

There are four operational modes available:

  • Off: The VPA calculates recommendations for CPU and memory based on historical data but does not automatically apply them to the Pods.^[400-devops__06-Kubernetes__k8s-ithelp__Day27__README.md]
  • Initial: The VPA applies resource recommendations only when a Pod is created.^[400-devops__06-Kubernetes__k8s-ithelp__Day27__README.md] Once the Pod is running, the VPA will not update its resources.^[400-devops__06-Kubernetes__k8s-ithelp__Day27__README.md]
  • Auto: The VPA automatically applies recommendations provided by the Recommender component.^[400-devops__06-Kubernetes__k8s-ithelp__Day27__README.md] In this mode, the Updater component actively evicts Pods that require updates, triggering the Admission Controller to patch the requests/limits before the Pod is recreated.^[400-devops__06-Kubernetes__k8s-ithelp__Day27__README.md]
  • Recreate: Similar to Auto, this mode automatically updates resources, but it forces the Pod to restart on every update cycle.^[400-devops__06-Kubernetes__k8s-ithelp__Day27__README.md] This mode is noted as being rarely used in practice.^[400-devops__06-Kubernetes__k8s-ithelp__Day27__README.md]

Sources