본문 바로가기

CKA &. CKAD/Design and Install a Kubernetes Cluster

Design an Kubernetes Cluster

Ask

ㅁ Purpose

  - Education

  - Development & Testing

  - Hosting Production Applications

ㅁ Cloud or OnPrem?

ㅁ Workloads

  - How many?

  - What kind?

     + Web

     + Big Data/Analytics

  - Application Resource Requirements

     + CPU Intetnsive

     + Memory Intensive

  - Traffic

     + Heavy Traffic

     + Burst Traffic

 


Purpose

 

  - Education

     + Minikube

     + Single node cluster with kubeadm/GCP/AWS

 

  - Development & Testing

     + Multi-node cluster with a Single Master and Multiple workers

     + Setup using kubeadm tool or quick provision on GCP or AWS or AKS

 

  - Hosting Production Application

    ?

 


Hosting Production Applications

 

  - High Availbility Multi Node cluster with multiple master nodes

  - Kubeadm or GCP or Kops on AWS or other supported platforms

  - Upto 5000 nodes

  - Upto 150,000 PODs in the cluster

  - Upto 300,000 Total Containers

  - Upto 100 PODs per Node

 

Nodes GCP AWS
1-5 N1-standard-1 1 vCPI 3.75GB M3.medium 1vCPU 3.75GB
6-10 N1-standard-2 2 vCPU 7.5GB M3.large 2vCU 7.5GB
11-100 N1-standard-4 4 vCPU 15GB M3.xlarge 4 vCPU 15GB
101-250 N1-standard-8 8 vCPU 30GB  M3.2xlarge 8 vCPU 30GB 
251-500 N1-standard-16 16 vCPU 60G C4.4xlarge 16 vCPU 30GB
> 500 N1-standard-32 32 vCPU 120GB C4.8xlarge 36 vCPU 60GB

 


Cloud or OnPrem?

 

  - Use Kubeadm for on-prem

  - GKE for GCP

  - EKS for AWS

  - AKS for Azure

 


Storage

 

  - High Performance - SSD Backed Storage

  - Multiple Concurrent connections - Network based storage

  - Persistent shared volumes for shared access across multiple PODs

  - Label nodes with specific disk types

  - Use Node Selectors to assign applications to nodes with specific disk types

 


Nodes

 

  - Virtual or Physical Machines

  - Minimum of 4 Node Cluster (Size based on workload)

  - Master vs Worker Nodes

  - Linux x86_64 Architecture

 

  - Master nodes can host workloads

  - Best practice is to not host workloads on Master nodes

 


Master Nodes