Node Pool Overview

Introduction

CCE introduces node pools to help you better manage nodes in Kubernetes clusters. A node pool contains one node or a group of nodes with identical configurations in a cluster.

You can create custom node pools on the CCE console. With node pools, you can quickly create, manage, and destroy nodes without affecting the cluster. All nodes in a custom node pool share the same type and configurations. You cannot configure a single node in a node pool. Any change applies to every node in the node pool.

You can also use node pools for auto scaling (supported only by pay-per-use node pools).

• When a pod in a cluster cannot be scheduled due to insufficient resources, scale-out can be automatically triggered.
• When there is an idle node or a monitoring metric threshold is met, scale-in can be automatically triggered.

This section describes how node pools work in CCE and how to create and manage node pools.

Node Pool Architecture

Figure 1 Overall architecture of a node pool

All nodes in a pool typically share:

• Node OS
• Node login mode
• Node container runtime
• Enterprise project
• Startup parameters of Kubernetes components on a node
• Custom startup script of a node
• Kubernetes labels and taints

CCE provides the following extended attributes for node pools:

• Node pool OS
• Maximum number of pods on each node in a node pool

Notes on a Node Pool Upgrade

For clusters of v1.21.11-r0, v1.23.9-r0, v1.25.4-r0, or later versions, new node pools support the creation of both pay-per-use and yearly/monthly nodes by default, providing you with better resource management experience.

The highlights of new node pools are as follows:

• Both pay-per-use and yearly/monthly nodes can be created in one node pool, and the required duration of yearly/monthly nodes can be different.
• Pay-per-use and yearly/monthly nodes can be accepted for management.
• Auto scaling can be enabled and various scaling policies can be configured for more efficient, flexible resource management.

The following table lists the changes after an existing node pool is upgraded.

Description of DefaultPool

DefaultPool is not a real node pool. It only classifies nodes that are not in the custom node pools. These nodes are directly created on the console or by calling APIs. DefaultPool does not support any user-created node pool functions, including scaling and parameter configuration. DefaultPool cannot be edited, deleted, expanded, or auto scaled, and nodes in it cannot be migrated.

Application Scenarios

When a large-scale cluster is required, you are advised to use node pools to manage nodes.

The following table describes multiple scenarios of large-scale cluster management and the functions of node pools in each scenario.

Functions and Precautions

Deploying a Workload in a Specified Node Pool

All nodes within a node pool carry the cce.cloud.com/cce-nodepool label. To ensure that a workload is scheduled onto nodes from a specific node pool, you can use the nodeSelector field in the workload settings. An example is as follows:

apiVersion: apps/v1
kind: Deployment
metadata:
  name: nginx
spec:
  replicas: 3
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      nodeSelector: 
        cce.cloud.com/cce-nodepool: "nodepool_name" # The label value is the node pool name.
      containers:
      - image: nginx:latest
        imagePullPolicy: IfNotPresent
        name: nginx
      imagePullSecrets:
      - name: default-secret

For more complex scheduling, you can define custom affinity rules, such as hard constraints, where scheduling occurs only if all specified conditions are met, and soft constraints, where scheduling may proceed even if some conditions are not met. For details, see Configuring Node Affinity Scheduling (nodeAffinity).

Additionally, you can specify resource requests for containers to ensure workloads are scheduled only on nodes that meet the required resource criteria. For details, see Resource Management for Pods and Containers. For example, if a workload pod requests four CPU cores, it will not be scheduled on a node that offers only two.

Helpful Links

You can log in to the CCE console and refer to the following sections to perform operations on node pools:

• Creating a Node Pool
• Managing Node Pools
• Creating a Deployment
• Configuring Node Affinity Scheduling (nodeAffinity)