What is K8S

• Open-source container orchestration tool
• Developed by Google
• With Increased usage of containers, demanding for a proper way of managing those hundreds of containers

Features

• High Availability
• Salability
• Disaster Recovery → Backup and restore

K8S Components

Node

• Simple a server, a physical machine, or a VM

Pod

• The smallest unit of K8S
• Abstraction over container
• Pods are meant to run only 1 container within it, 1 app/pod
• Each pod is assigned an IP address
• Pods communicate with each others using IP address
• Pods can die easily → When pods die, they got assigned a new IP on creation

Service

• Permanent IP address that can be attached to each pod
• Lifecycle of Pod and Service are not connected
• External Service: Exposed its communication to external sources
• Internal Service: Not exposed to the world

Ingress

• Forward requests to Services

Config Map

• external configuration of the application

Secrets

• Used to store secret data
• Base64 encoded

Volumes

• Attaches a storage onto the pod
• The storage can be local or remote (outside of the K8S cluster like cloud or on-premise)
• K8S doesn’t manage data persistence

Deployment

• blueprint for my-app pods
• Database can’t be replicate via deployment

StatefulSet

• Deploys for stateful Apps or Databases

K8S Architecture

Worker Machine in K8S cluster

• Each Node has multiple Pods on it
• 3 processes must be installed on every Node
• Worker Nodes do the actual work

Container runtime (Docker) is installed on all nodes

• kubelet
  • Processes that schedules the pods
  • It interacts with both the container and node
• Nodes are communicate via Services
• Kube Proxy:
  • Must be installed on every Nodes
  • Forwards the requests to the nodes
  • Making sure that the performance is performing with the lowest overheads

Master Processes

• Have 4 processed that run on every master node
  1. API Server
     • Cluster Gateway
     • Client interact with the API Server to create/manage new applications
     • Act as a validator which authenticate the Client Requests
  2. Scheduler
     • Decide which of the worker node to put the pod in
     • See how much resources the application that client want to schedule will need (how much CPU, RAM)
     • Choose the Node to schedule it in via Kubelet
  3. Controller Manager
     • When a pod dies on any Node, The Controller Manager request the Scheduler to recover the pods
  4. etcd
     • The cluster Brain.
     • Cluster changes get stored in the key value store.
• Master Nodes consumes less resources (CPU, RAM, Storage)

Example Cluster Set-Up

Add new Master/Node server

1. Get new bare server
2. Install all the master/worker node processes
3. Add it to the cluster