Skip to content

1.5 - Cluster API

Objectives

  • Describe Cluster API
  • List infrastructure providers
  • Detail Cluster API controllers
  • List the Cluster API Custom Resource definitions

Introduction

  • Cluster API is a Kubernetes project to provide declarative Kubernetes-styled APIs to cluster provisioning, configuration and management
  • Cluster API controllers running on a Kubernetes cluster receives Cluster API definitions that specify the desired state of the new cluster
  • They also send requests to the cloud/IaaS provider(s) to create the new cluster by using the required provider configurations.

Untitled

CRD = Custom Resource Definition

Cluster API Custom Resource Definitions (CRDs)

  • Cluster API provides custom resource definitions (CRDs) to management clusters to help the management cluster know what to deploy to the particular workload cluster(s).

Untitled

Machine Health Checks

  • Allows cluster API to monitor the health of cluster nodes
  • A timeout is specified for each of the conditions that can be checked
  • If any of the conditions are met for the duration of the timeout, the machine is remediated
  • The action of remediating a machine triggers a new machine to be created to replace the old one.

Untitled

  • **Note: This is equivalent to the traditional Readiness and Liveness probe functionalities available in Kubernetes for deployments.

Cluster API Components

  • Logic in the Cluster API is divided per component
  • Bootstrap providers - perform the steps to configure and start the Kubernetes processes that make up a cluster
  • Infrastructure providers - perform the steps to create cloud resources e.g. VMs
  • Basically like how providers work in Terraform

Bootstrap Provideers

  • Responsible for turning a machine into a Kubernetes Node
  • Cluster API bootstrap provider kubeadm is a bootstrap provider implementation that utilises kubeadm to perform machine configuration to set up a Kubernetes Cluster (management in this case).

Infrastructure Providers

  • All in format CAP
  • E.g. CAPV = vSphere, CAPA = AWS, CAPD = Docker
  • Only some of these are supported by Tanzu Kubernetes Grid 1.3 directly:
  • CAPV - vSPhere
  • CAPA - AWS
  • CAPZ - Azure

Cluster API Controllers

  • Responsible for watching the Kubernetes API for cluster API (CAPI) resources and perform the required steps to provision clusters

Untitled

Note: This is analagous to the traditional kube-system deployments like kube-scheduler, kube-controller-manager, etc.

vSphere Cluster API Provider (CAPV)

  • An infrastructure provider that provides the following controller and custom resource definitions

Untitled

Summary

  • Note: Cluster API ONLY runs on the Management Cluster

  • Note: You can think of them equivalent to the traditional kube-system deployments