Node Problem Detector (NPD) is deployed as a daemon set which runs on worker and Infra nodes. It is designed to inspect various subsystems and components of the node and report any problems or anomalies it detects to apiserver. It aims to make various node problems visible to the upstream layers in the cluster management stack, this information helps Cluster Administrators proactively identify and address issues that could impact the node’s health and stability.

A descheduler is a specialized component in Kubernetes that helps to improve cluster resource utilization and balance by evicting (and subsequently scheduling elsewhere) Pods from nodes that meet certain criteria. It operates as a daemon alongside the regular Kubernetes scheduler.