DevOps Classroom notes 03/Oct/2026

Kubernetes Architecture Overview

A Kubernetes cluster has two parts:

1. Control Plane (master): the brain. Makes decisions and stores cluster state.
2. Worker Nodes: the muscle. Run your application containers.

flowchart TB
    User["User / kubectl / CI-CD"] -->|HTTPS REST| API

    subgraph CP["Control Plane"]
        API["kube-apiserver"]
        ETCD[("etcd")]
        SCHED["kube-scheduler"]
        CM["kube-controller-manager"]
        CCM["cloud-controller-manager"]
        API <--> ETCD
        SCHED --> API
        CM --> API
        CCM --> API
    end

    subgraph W1["Worker Node 1"]
        K1["kubelet"]
        P1["kube-proxy"]
        R1["Container Runtime"]
        POD1["Pods"]
        K1 --> R1 --> POD1
    end

    subgraph W2["Worker Node 2"]
        K2["kubelet"]
        P2["kube-proxy"]
        R2["Container Runtime"]
        POD2["Pods"]
        K2 --> R2 --> POD2
    end

    API <--> K1
    API <--> K2

Master componets:

Kube api-server:

• this responsible for communtion etcd.
• speak with api server by using kubectl
• this end point to access cluster via restapi over https
• this will handle auth methods, RBAC…etc
• stateless, so you can run multiple replicas

etcd

• data will store in etcd
• it distubution store type have data store in key-value pair
• back of cluster we need take snapshort etcd

metadata: 
  - name: fronent
    kind: pod
    container:
        - name: webapp
          image: nginx
          tag: latest
          volume: datasource
          network: default
          port: 80
  - name: fronent
    kind: deployment
    container:
        - name: webapp
          image: nginx
          tag: latest
          volume: datasource
          network: default
          port: 80
        - name: webapp
          image: nginx
          tag: latest
          volume: datasource
          network: default
          port: 80

kube-scheduler:

• watches a new schedule pods in k8s nodes
• monitoring pods (cpu, memory, storage, networkwork , affinity)

kube-controller

• node control
• job control
• cronjob –> cronexpression –> job –> pod created
• replicas/deploymnet controller
• service account / Token controller

colud controller-manager

• this runing on only cloud managed clusters (EKS, AKS, GKE)
• it managed k8s apiserver by cloud-api provider , loadbalancer and routes
• not present local steps like kubadm, kind, minikube, k8s bare-meteal, rke2

Worker node componets:

kubelet

• this is agent which is runing in nodes
• watch pods and asign node to run container
• this agnet will speak and do work by api server inputs
• manage servers deployed in worker node

container Runtime

• software that runs container and pull images
• must be implement CRI

kube proxy:

• runs on every node
• implement the service networking
• manage ip address, tls rules , traffic , communtion.

Steps to install kubeadm in ubuntu