What is the function of Horizontal Pod Autoscalers (HPA) in Kubernetes?

What is the function of Horizontal Pod Autoscalers (HPA) in Kubernetes?

• A. They scale the resource requests and limits of a pod
• B. They scale the number of replicas for an application
• C. They adjust the size of the Kubernetes Cluster
• D. They enforce quotas on namespaces

Answer: (B)

The key idea is automatically adjusting the number of pod replicas for a workload based on observed metrics. Horizontal Pod Autoscaler watches metrics such as CPU utilization (and can use other custom metrics) and updates the replica count of a target workload (like a Deployment or ReplicaSet) to keep the chosen target metric value near the desired level. If the workload is under heavy load and metrics are above the target, the HPA scales out by adding more pods; if the load drops and metrics are below the target, it scales in by reducing pods. This scaling is bounded by configured minimum and maximum replicas. It does not change the resource requests or limits of individual pods, nor does it resize the cluster itself. Adjusting per-pod resources is done by altering the pod spec or using other mechanisms, while scaling the cluster depends on components like a Cluster Autoscaler. It also doesn’t enforce namespace quotas—that’s the job of quota and limit enforcement in Kubernetes.

The key idea is automatically adjusting the number of pod replicas for a workload based on observed metrics. Horizontal Pod Autoscaler watches metrics such as CPU utilization (and can use other custom metrics) and updates the replica count of a target workload (like a Deployment or ReplicaSet) to keep the chosen target metric value near the desired level. If the workload is under heavy load and metrics are above the target, the HPA scales out by adding more pods; if the load drops and metrics are below the target, it scales in by reducing pods. This scaling is bounded by configured minimum and maximum replicas.

It does not change the resource requests or limits of individual pods, nor does it resize the cluster itself. Adjusting per-pod resources is done by altering the pod spec or using other mechanisms, while scaling the cluster depends on components like a Cluster Autoscaler. It also doesn’t enforce namespace quotas—that’s the job of quota and limit enforcement in Kubernetes.