강의영상 : https://www.youtube.com/watch?v=s_o8dwzRlu4

1. What is Kubernates?

Definition

• open source container orchestration tool(e.g., docker contatiner)

• Developed by Google

• Helps manage containerized applications in different deployment environments(physical machine/virtual machine/cloud/hybrid)

Need for container orchestration tool

• Trend from Monolith to Micorservices

• Increased usage of contatiners

• Demand for a proper way of managing those hundreds of containers

Features that orchestration tools offer

• High Availability or no downtime

• Scalability (load - performance)

• Disaster recovery (back up and restore)

2. Kubernetes Architecture

Nodes and Pod

• at least one Master node(control plane) + a couple of worker nodes(=nodes)

Master node

• have to at least two Master nodes inside of Kubernates cluster because it is important

• runs several Kubernetes processes that are absoultely necessary to run and manage the cluster properly

1) API server : Entrypoint to K8s cluster(process which different kubernetes clients talk to API server, e.g., UI(dashboard), API, Command line tool(KUBECTL) : YAML or JSON)

2) Controller Manager : Keeps track of what happening in the cluster(whether something nees to be repaired or maybe if a container died and it needs to be restarted etc)

3) Scheduler : decides on which Node new Pod shold be scheduled based one the available resources on each worker node and the load that that container needs

4) etcd : Kubernetes backing store(key value storage) which basically holds at any time the current state of the Kubernetes cluster, e.g., configuration data, status data of each node and each container. Backup and restore are made from etcd snapshots

Worker node

• runs actual work

• Each node has a Kubelet(node agent) process running on it

• Kubelet is a Kubernetes process that maskes it possible for the cluster to communicate each other and execute some task on those nodes

• high load

Virtual Network

• enables master nodes and worker nodes talk to each other

• creates one unified machine

Pod

• smalles unit in Kubernetes

• abstraction over container

• creates running environment or a layer on top of the container

• Pod(not the container) gets its own IP address
  : my application container can communicate with database container using the IP address)

• Pods are ephemeral(easy to die) : need Permanent IP address by Service(another kubernetes component)

• Usually 1 application per Pod(or multiple help/subservice applications)

• my-app Pod, DB Pod, etc

Service & Ingress

Service

• Permanent IP address

• Lifecycle of Pod and Service not connected

• You specify the type of Service on creation
  (External/Internal-default)

• You have http protocol with a node IP address and the port number of the Service (e.g., http:/124.89.101.2:8080)

• Load balancer : catch the request and forward to whichever part is least busy

• One Service - multiple Pod(to prevent down time)

Ingress

• https + domain name

• request goes first to Ingress and it does the forwarding them to the Service

ConfigMap & Secret

• used as environment variables or a properties file

ConfigMap

• External configuration of application

• urls of a database or some other services

• By ConfigMap, if you change the nane of service/the endpoint of the service you just adjust config map(don't have to build a new image and push it to repo and pull it in your Pod)

• for non-confidential data only

Secret

• is just like ConfigMap, but used for secret data(e.g., credentials)

• stored in base 64 in encoded format

• need third-party tools to encrypt them

Volume

• for reliably long term persistance of database/log data

• storage on local machine, or remote(outside of the K8s cluster) machine

• Kubernetes doesn't manage data persistance(You have to backup)

Deployment & StatefulSet

Deployment

• blueprint for Pods

• set the number of replicas(one Service - multiple Pods), scaling, etc

• Abstraction of Pods

• not for database pods

StatefulSet

• for stateful apps(e.g., mysql, mongoDB, elastic search

• database reads and writes should be synchronized to prevent data inconsistencies

• deploying StatefulSet in Kubernetes cluster is not easy -> DB are often hosted outside of Kubernetes cluster

3. Kubernetes Configuration

• To API server by UI/API/CLI(YAML/JSON)

• Declarative(Controller Manager checks : desired state == actual state?)

• It goes to work to make sure that this desired state is recovered automatically restarting the second replica of that Pod

3 Parts of a K8s Configuration File

• apiVersion

• kind : Component type

1) metadata

• name of component, etc

2) specification

• are specific to the kind

3) status

• automatically generated and edited by Kuber netes

• Kubernetes always compare what is the desired state and what is the actual state(by etcd)

• if the two states do not match, Kubernates try to fix some difference(Self-healing features)

YAML

• human friendly data serialization standard for all programming languages

• syntax : stric indentation

• store the config file with your applictaion code

• or its own git repository

4. Minikube & Kubectl - Local Setup

Minikube

Production Cluster Setup

1) multiple Master and Worker nodes

2) separate virtual or physical machines

=> testing on local machine is difficult and maybe impossible if you don't have enough resources(need Minikube)

Minikube

• open source tool

• one Node(machine) cluster

• both the master processes and worker processes run on one node

• this node will have a docker container runtime pre-installed

Kubectl

• command line tool for K8s cluster(API server)

• The most powerful of 3 clients(UI?API?CLI)

• You can do anything to Kubernetes, e.g., creating Pods, destroying Pods, creating Services