Cloud-Native Microservices With Kubernetes - 2nd Edition

PriorityClass

A PriorityClass is a cluster-wide (non-namespaced) resource from the scheduling.k8s.io API group. It defines a mapping between a priority name and an integer value. Higher numbers indicate higher priority.

This is an example:

apiVersion: scheduling.k8s.io/v1
kind: PriorityClass
metadata:
  name: high-priority
value: 1000000
globalDefault: false
description: "Used for critical Pods such as monitoring agents"
preemptionPolicy: PreemptLowerPriority

The fields are:

• value — An integer between -2147483648 and 1000000000. Higher values represent higher priority. Values above 1 billion are reserved for Kubernetes system Pods.

• globalDefault — When set to true, this PriorityClass becomes the default for all Pods that do not explicitly define a priority. Only one PriorityClass in the cluster can have this field set.

• description — A human-readable note explaining when to use this PriorityClass.

• preemptionPolicy — Determines whether Pods in this class can preempt lower-priority Pods. The options are:

• PreemptLowerPriority (default): Allows eviction of lower-priority Pods.

• Never: Pods with this class cannot preempt others, but they are still scheduled before lower-priority Pods once resources are available.

ℹ️ In multi-tenant clusters, a malicious user could create Pods with very high priorities and evict others. Cluster administrators can prevent this using ResourceQuotas that restrict access to specific PriorityClasses.

Non-preempting PriorityClass

Starting with Kubernetes v1.24, you can create non-preempting PriorityClasses. These Pods wait in the scheduling queue ahead of lower-priority Pods, but never evict them.

Example:

apiVersion