Cluster

What is a Cluster?

• When you deploy Kubernetes, you get a Cluster

• A Kubernetes Cluster is a set of worker machines called nodes that run containerized applications

• Every cluster has at least one worker node

Cluster Architecture

At the hardware level, a Kubernetes Cluster consists of multiple nodes, which can be divided into two types

1. Master Node

   Runs the Kubernetes Control Plane which controls and manages the entire Kubernetes system

2. Worker Node

   Runs the actual deployed container applications

Control Plane

The Control Plane controls and operates the cluster

It consists of the following components that can run on a single master node or be split and replicated across multiple nodes to ensure high availability

Kubernetes API Server

The API Server is the component that exposes the Kubernetes API

Communicates with users and Control Plane components

kube-apiserver

• kube-apiserver is designed to scale horizontally

  • That is, it can be scaled by deploying more instances

• You can run multiple kube-apiserver instances and balance the traffic between them

Scheduler

A component that detects newly created Pods with no assigned node and selects a node to run them on

• Responsible for the deployment of applications

  • Assigns each deployable component of an application to a worker node

• Factors considered for scheduling include

  • Individual/aggregate resource requirements

  • Hardware/Software policy constraints

  • Affinity & anti-affinity specifications

  • Data locality

  • Inter-workload interference

  • Deadlines

Controller Manager

A component that runs controller processes

• Performs cluster-level functions such as replicating components, tracking worker nodes, and handling node failures

• Logically, each controller is a separate process, but to reduce complexity, they are compiled into a single binary and run within a single process

Controller Types

• Node controller

  • Responsible for notifying and responding when a node goes down

• Job controller

  • Watches for one-time Job objects and creates Pods to execute those jobs

• Endpoints controller

  • Connects Services and Pods

• Service Account & Token controllers

  • Creates default accounts and API access tokens for new Namespaces

etcd

A reliable, key-value structured distributed data store that continuously stores the cluster configuration

The components of the Control Plane maintain and control the cluster state, but they do not run applications. That is done on the Nodes.

Node

Worker Nodes are systems that run containerized applications

Running and monitoring applications, and providing services to applications, is performed by the following components

Container Runtime

Responsible for running containers

Supports Docker, containerd, CRI-O, or any software that implements the Kubernetes CRI (Container Runtime Interface)

Kubelet

Communicates with the API Server and manages containers on the node

• An agent that runs on each node in the cluster, managing containers to ensure they run within Pods

• Receives a set of PodSpecs through various mechanisms and manages containers to ensure they are running properly according to those Pod specs

• Does not manage containers that were not created through Kubernetes

Kube-proxy

Kube-Proxy load-balances network traffic between application components

• A network proxy that runs on each node in the cluster, implementing the Kubernetes service concept

• Maintains and manages network rules on the node

  • These network rules allow network communication to pods from internal network sessions or from outside the cluster