The automatic scale out feature works based on the performance metrics (metrics) of objects in a Kubernetes cluster. Metrics are provided by a service that implements the metrics API, called a metrics server.

There are three types of metrics servers (metrics APIs) in OpenShift/Kubernetes.

• Standard metrics server

• Custom metrics server

• External metrics server

Automatic scale out based on CPU utilization works with metrics obtained from standard metrics server.
Automatic scale out based on the number of connections works with metrics obtained from custom metrics server.

The custom metrics server looks at the metrics information for the FEP cluster collected by the monitoring function in Prometheus and publishes it in the form of a metrics API.

Custom metrics server are resources shared by OpenShift/Kubernetes clusters, so they are not built and configured for an extended FEP for Kubernetes installation. To take advantage of automatic scale out based on the number of connections, you must have a custom metrics server.

The custom metrics server must also reference Prometheus, which collects custom metrics for FEP clusters, and must be configured to publish custom metrics for FEP clusters.