Design And Implementation Of A Highly Available Container Cloud Based On Kubernetes

With the rise of container technology and Kubernetes,many cloud vendors around the world have launched their own container cloud services.The wave of container technology started with the popularity of Docker,and only with the gradual maturation of Kubernetes did it really take hold,and only then did it start to truly transform from the traditional application deployment model to the container service model.80% of Ali's core business has been deployed to the container cloud platform during the 2020 double11,and Google has accumulated decades of container services experience in the production environment,Kubernetes is its open source container orchestration project.Kubernetes sits on top of the cloud Iaa S infrastructure layer,shielding the underlying hardware from complexity and providing users with automated deployment,scaling and management of containerized applications.The rise of kubernetes is precisely because the traditional application architecture is changing to the microservice architecture,and kubernetes just fits the ideas and best practices of microservice and Devops.It is a challenge for many enterprises to migrate traditional applications to Kubernetes container cloud platform.And how to build a Kubernetes-based container cloud platform has become a popular research direction.Although the open source Kubernetes provides powerful container orchestration and management functions,it is not perfect in high availability,multi-tenancy and application lifecycle management.This thesis designs and implements a Kubernetes-based highly available container cloud platform on top of the lab's existing Iaa S tier cloud platform,providing highly available container cloud services,unified multi-tenancy management,and full lifecycle application and container image management.This thesis mainly designs and implements the high availability of container cloud services from three dimensions.Firstly,the etcd storage management module is designed in the storage dimension to provide health check and fault recovery for external etcd clusters to enhance the high availability of the storage layer.Next,a high availability solution based on a combination of Keepalived and HAProxy and a multi-control node deployment topology is proposed at the control node level.Thirdly,at the working node level,a highly available elastic scaling module is designed and implemented to provide automatic elastic scaling of Kubernetes cluster nodes,and a scheduler is implemented based on the Kubernetes scheduler framework for secondary scheduling optimization of Pods.In addition,this thesis also designs a unified multi-tenancy module to introduce multi-tenancy for native Kubernetes,which implements Open ID Connect-based open unified authentication and user role-based access control.Finally,from the perspective of system implementation integrity,the application and container image management module is implemented to achieve the full lifecycle management of Kubernetes applications and images.