HPA Dynamic Scaling for CN Nodes with Helm Charts
This document describes how to implement Horizontal Pod Autoscaler (HPA) based dynamic scaling for CN (Compute Node) nodes in PhoenixAI clusters using Helm Charts. This feature was introduced in v1.11.0 and addresses critical autoscaling issues including resource cleanup, version compatibility, and graceful scaling operations.
Scale in FE nodes
FE nodes in a PhoenixAI cluster are used to store metadata. Normally, users do not need to scale in the FE nodes.