Kubernetes Cluster Architecture

 

A Kubernetes Cluster contains:

┌─────────────────────────────┐

│ Control Plane         │

└─────────────────────────────┘

|

------------------------------------------

| | |

Worker1 Worker2 Worker3

Control Plane (Master Node)

Brain of Kubernetes.

Responsible for:

Scheduling

Monitoring

Cluster decisions

State management

Cluster 

 ├── Control Plane

 │ ├── API Server 

 │ ├── ETCD

 │ ├── Scheduler

 │ └── Controller Manager

 │

 └── Worker Nodes

    ├── Kubelet

    ├── Kube-Proxy

    └── Container Runtime

Components:

API Server

kubectl --> API Server --> Cluster

Entry point of Kubernetes.

Every request goes through API Server.

Examples:

kubectl get pods

kubectl create deployment nginx

ETCD

Database of Kubernetes.

Stores:

Cluster state

Pod information

Secrets

ConfigMaps

Node info

Example:

Current State: 3 Pods running

Stored inside ETCD.

Scheduler

Decides: Which worker node should run a pod?

Example:

Cluster:

Node1 CPU = 90%

Node2 CPU = 20%

Scheduler chooses:

Pod --> Node2

because Node2 has more resources.

Controller Manager

Monitors desired state versus current state.

Example:

Desired: 3 Pods

Current: 2 Pods

Controller notices mismatch and starts: New Pod TO MAKE IT 3 Pods

== Worker Node Components == 

Worker node is actual responsible for manage and runs your applications.

Kubelet

Agent running on every worker node.

Responsibilities:

Talks with API Server

Creates pods

Health checking

Think:

Manager --> Worker

API Server --> Kubelet

Container Runtime

Responsible for container execution.

Examples:

containerd

CRI-O

Older Kubernetes: Docker

Kube-Proxy

Handles network traffic.

Responsibilities:

Service routing

Load balancing

Example:

User Request 

|

Kube-Proxy

|

------------------

|       |        |

Pod1   Pod2    Pod3