Introduction, architecture and principle of kubernetes

Official website document address ：https://kubernetes.io/zh/docs/home/ （k8s There are official detailed documents , The best way to learn is to go to the official website ）

kubernetes brief introduction

Kubernetes It is an automatic container arrangement platform , It is responsible for the deployment of applications 、 Application flexibility and application management . It can provide many functions , Generally speaking kubernetes It is a container cloud steward , It manages and orchestrates containers ,kubernetes The following functions can be provided :

Service discovery and load balancing

Kubernetes have access to DNS Name or own IP Address disclosure container , If the flow into the container is large , Kubernetes Can load balance and distribute network traffic , So that the deployment is stable

Storage choreography

Kubernetes Allows you to automatically mount the storage system of your choice , For example, local storage 、 Public cloud providers, etc .

Automatic deployment and rollback

You can use Kubernetes Describes the required state of the deployed container , It can transfer the actual state at a controlled rate Change to the desired state . for example , You can automate Kubernetes To create a new container for your deployment , Delete existing containers and use all their resources for new containers .

Automatically complete the packing calculation

Kubernetes Allows you to specify what each container needs CPU And memory （RAM）. When the container specifies a resource request ,Kubernetes Better decisions can be made to manage the container's resources .

Self repair

Kubernetes Restart the failed container 、 Replace the container 、 Kill does not respond to user-defined Health check container , And don't advertise the service to the client until it's ready .

Key and configuration management

Kubernetes Allows you to store and manage sensitive information , For example, password 、OAuth Token and ssh secret key . You can deploy and update the key and application configuration without rebuilding the container image , There is no need to expose the key in the stack configuration

… etc.

kubernetes Cluster architecture

kubernetes It uses master-node framework .
Master As the central control node , Will go with Node Make a connection . client （ such as UI/CLI etc. ） Only with Master Connect , Send the desired state or order to Master,Master These commands or states will be sent to the corresponding nodes , Carry out the final execution

• master: Master node , There may be a lot of
• node: Work node , quite a lot , The application of real work

working principle

Master

Kubernetes Of Master There are four main components ：kube-apiserver、kube-controller、kube-scheduler as well as etcd, As shown in the figure below . We usually call it Master The node is the control surface （Control Plane）, This of the control surface 4 Components make global decisions about the cluster （ For example Pod Schedule to a suitable node ）, And detecting and responding to cluster Events .

master There are some core components on the node ：

• Controller Manager： Control Manager , It is used to manage the cluster state
• etcd： Key value database , It's a distributed storage system ,API Server All the original information needed in is placed in etcd in ,etcd Itself is a highly available system , adopt etcd Guarantee the whole Kubernetes Of Master High availability of components
• scheduler： Scheduler ,“ Scheduler ” As the name implies, it is to complete the scheduling operation , This component monitors those newly created Pod, And select the node that conforms to the regulation to let Pod Run on . The factors considered in scheduling decision include single Pod and Pod Aggregate resource requirements 、 Hardware / Software / Policy constraints 、 Affinity and anti affinity norms 、 Data location 、 Interference between workloads, etc .
• api server：api gateway , As the name suggests, it is used to deal with API Operation of the ,Kubernetes All of the components in will be associated with API Server Connect , There is no independent connection between components , Rely on a API Server To transmit messages

Node

Kubernetes Of Node It's really running the business load , Each business load will be loaded with Pod Form operation . One Pod Run in one or more containers , To really run these Pod The component of is called kubelet, That is to say Node The most critical component on , It passes through API Server Receive what you need Pod State of operation , And then submit this Container Runtime Components （ A simple understanding is what we usually call a container ） in .

node node （worker Work node ）：

• kubelet： This is the most important component running on each work node ,kubelet Receive a set of PodSpecs（ That is to say Pod The expected state of ）, Make sure these PodSpecs The container described in is running and healthy .
• kube-proxy： agent . Agent network ,kube-proxy Maintain network rules on nodes . These network rules allow you to interact with Pod Network communication . It is the realization of Kubernetes Service Important components of the concept .
• Pod: Pod yes Kubernetes A minimum scheduling and resource unit of . The user can go through Kubernetes Of Pod API Produce a Pod, Give Way Kubernetes For this Pod To schedule , That is to put it in a certain Kubernetes Running on managed nodes . One Pod It's an abstraction of a set of containers , It will contain one or more containers .

Master and node The node interaction is shown in the following figure ：

It's important to note that ,Kubernetes Of Node It will not interact directly with users , Its interaction will only be through Master, User pass Master Send information to the node .Kubernetes Every Node On , Will run the components mentioned just now .

Principle of component interaction

Want to make k8s Deploy a tomcat？ The whole interaction process is as follows ：

1. Power on defaults to all nodes kubelet、master Node scheduler（ Scheduler ）、controller-manager（ Control Manager ） Keep monitoring master Of api-server Changes in events （for ::）
2. Programmers use command-line tools ： kubectl ; kubectl create deploy tomcat --image=tomcat8（ tell master Let the cluster use tomcat8 Mirror image , Deploy a tomcat application ）
3. kubectl The command line content is sent to api-server,api-server Save the creation information to etcd
4. etcd to api-server Report the incident , Said someone just saved a message for me .（ Deploy Tomcat[deploy]）
5. controller-manager Listen to the api-server Events ,（ Deploy Tomcat[deploy]）
6. controller-manager Deal with this （ Deploy Tomcat[deploy]） Events .controller-manager Will generate Pod Deployment information for 【pod Information 】
7. controller-manager hold Pod Give your information to api-server, Save it to etcd
8. etcd Report the incident 【pod Information 】 to api-server.
9. scheduler Special monitoring 【pod Information 】 , Get 【pod Information 】 The content of , Calculation , See which node is suitable for deploying this Pod【pod Information after scheduling node: node-02）】
10. scheduler hold 【pod Information after scheduling （node: node-02）】 hand api-server Save to etcd
11. etcd Report the incident 【pod Information after scheduling （node: node-02）】, to api-server
12. Of other nodes kubelet Special monitoring 【pod Information after scheduling （node: node-02）】 event , Cluster all nodes kubelet from api-server Got it. 【pod Information after scheduling （node: node-02）】 event
13. For each node kubelet Judge whether it belongs to you ;node-02 Of kubelet It was his business to find out
14. node-02 Of kubelet Start this pod. Report to master All the information currently started