Which component increases or decreases the Kubernetes cluster size by adding or removing nodes based on usage?

Which component increases or decreases the Kubernetes cluster size by adding or removing nodes based on usage?

• A. Horizontal Pod Autoscaler
• B. Vertical Pod Autoscaler
• C. Keda ScaledObjects
• D. Cluster Autoscaler

Answer: (D)

Cluster size is adjusted by watching the cluster’s resource needs and provisioning or deprovisioning workers accordingly. The component that handles this across cloud providers by adding or removing nodes is the Cluster Autoscaler. It looks for pods that can’t be scheduled due to insufficient nodes and then requests more nodes to be added to the cluster. When there are nodes that aren’t needed—often because workloads have scaled down or are idle—the autoscaler can remove those nodes to save costs, respecting the defined min and max node counts. This is different from horizontal pod autoscaling, which only increases or decreases the number of replicas of individual pods within the existing cluster, not the number of nodes. Vertical pod autoscaling changes the resource requests and limits of containers in pods, again without changing the cluster size. KEDA scaled objects drive pod scaling based on external metrics or events, also within the current cluster capacity.

Cluster size is adjusted by watching the cluster’s resource needs and provisioning or deprovisioning workers accordingly. The component that handles this across cloud providers by adding or removing nodes is the Cluster Autoscaler. It looks for pods that can’t be scheduled due to insufficient nodes and then requests more nodes to be added to the cluster. When there are nodes that aren’t needed—often because workloads have scaled down or are idle—the autoscaler can remove those nodes to save costs, respecting the defined min and max node counts.

This is different from horizontal pod autoscaling, which only increases or decreases the number of replicas of individual pods within the existing cluster, not the number of nodes. Vertical pod autoscaling changes the resource requests and limits of containers in pods, again without changing the cluster size. KEDA scaled objects drive pod scaling based on external metrics or events, also within the current cluster capacity.