CKA를 준비하면서 공부한 요약 내용입니다.

• 강의
• What is CKA?

Design a k8s cluster

Purpose

• Education
  • Minikube
  • Single node cluster w\ kubeadm/GCP/A
• Development & Testing
  • Multi-node cluster with a single master and multiple workers
  • setup using kbeadm tool or quick provision on GCP or AWS or AKS
• Hosting Production Applications

Hosting Production Applications

• High Availability 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

Cloud or OnPrem

• Use Kubeadm for on-prem
• GKE for GCP
• Kops 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
• Master vs Worker Nodes
• Linux X86_64 Architecture
• Master nodes can host worloads
• Best practice is to not host workloads on Master nodes

ETCD in HA

ETCD

• a distributed reliable key-value store that is Simple, Secure & Fast

Key-value store

• Tablular/Relational Databases
• key-value store

Distributed

• it is possible to have your database across multiple servers

Consistent

• how does it ensure the data on all the nodes are consitent?
  • ETCD ensures that the same consistent copy of the data is available on all instances at the same time
• how does do that?
  • read ← easy
  • write
    • two write request coming
    • only one of the instances is responsible for processing the writes
    • interanally the other nodes elect a leader among them.
      • one node: leader
      • other nodes: followers
    • leader process write
      • leader makes sure that the other nodes are sent a copy of the data
      • if the writes come in through any of the other follower nodes, then they forward the wirtes to the leader internally

Leader Election - RAFT

• How to elect leader in ETCD
• → use RAFT protocol

방법

• use RAFT Algorithm
  1. random timers for initiating requets
  2. first nodes sent out a requetst to the otehr node requesting permission to be the leader
  3. other managers receiving the request responds with their vote and node assumens the leader role
  4. leader is elected, it sends out notifications at regular intervals to other masters informing them that it is continuing to assume the roloe of the leader
• other nodes do not receive a notifications from the leader at some point time
  1. nodes initiate a re-election process among themselves
  2. new leader is identified