Cloud-Native Microservices With Kubernetes - 2nd Edition

Node Reserve vs Allocatable Resources

In a Kubernetes cluster, each node reserves resources for the operating system and Kubernetes components. These resources are referred to as node reserve resources. The remaining resources are known as allocatable resources.

The node reserve resources are intended for components such as:

• The operating system
• The kubelet
• The kube-proxy
• The container runtime
• Node-level logging and monitoring agents
• Node-level network plugins

These resources are not to be used by Pods running on the node. Pods, on the other hand, have their own resources, termed allocatable resources. This diagram illustrates this concept:

Here is a list of the different types of node reserved resources:

• kube-reserved — Capacity carved out for Kubernetes system daemons that do not run as Pods on the node (e.g., kubelet, container runtime, CNI binaries). This reduces the node’s allocatable, so Pods can’t consume it. The values are set by the operator (they don’t auto-scale with Pod density, though you might size them based on expected load).

• system-reserved