摘  要

容器集群的web应用是指使用软件引擎如Docker在服务器集群的操作系统上部署软件容器，再在其中部署web应用程序的开发方案，它最初是由Docker公司在2013年设计开发的。由于Docker引擎提供了一个额外的软件抽象层，以及操作系统层虚拟化的自动管理机制，使得web应用的部署兼备灵活性和便携性，于是使用Docker容器集群的web应用逐渐成为业界的热门方案。为了使容器集群的web应用能够承载海量的数据和服务并发量，达到较高的可用性和可靠性，开发设计人员必须尽可能地使容器中的应用充分利用服务器资源。但是在实际操作中，开发设计人员由于需要通过繁琐的操作进行手动管理，往往难以观察到容器集群的实时性能，进而影响了负载不均问题的解决。所以，解决以上问题的最好办法是实现一个基于容器集群的web应用性能的自动化分布式测试工具来处理容器集群的性能问题。

本文研究内容包括：容器集群应用的部署和测试工具、web应用性能的检测、性能数据的可视化表示。Docker容器集群的应用有其独特的体系架构，课题需要研究Kubernetes集群配置的过程和原理，比如包括部署Ectd集群、Kucectl命令行工具、Flannel 网络等，并优化部署分配模块，优化测试页面结构，使用户可以通过简易操作实现自动化分布式部署和测试。最后，实现集群部署和测试工具的用户交互界面，性能数据导出后生成可视化界面，提供直观形象的集群性能反馈。

关键字：容器；集群；Web；分布式；Kubernetes

Abstract

Container cluster web applications refer to the deployment of software containers on server cluster operating systems using software engines such as Docker, which was originally designed and developed by Docker in 2013.Because the Docker engine provides an additional software abstraction layer, as well as an automated management mechanism for operating system layer virtualization, web applications are deployed with both flexibility and portability.Therefore, the web application using Docker container cluster has gradually become a hot solution in the industry.To enable container cluster web applications to carry the sea Since the amount of data and service concurrency can achieve high availability and reliability, developers and designers must make full use of server resources in the container as far as possible.However, in practice, developers and designers often find it difficult to observe the real-time performance of container clusters because of the need for manual management through cumbersome operations, thus affecting the solution of the problem of uneven load.Therefore, the best way to solve the above problem is to implement an automated distributed testing tool for web application performance based on container cluster to deal with the performance problems of container cluster.

The research contents of this paper include: the deployment and testing tool of container cluster application, the visual representation of performance data, the application of Docker container cluster has its unique architecture.We need to study the process and principle of Kubernetes cluster configuration, such as deploying ECTD cluster Kucectl command-line tool and Flannel network, optimizing deployment allocation module, optimizing test page structure, etc.Enables users to automate distributed deployment and testing through simple operations.Finally, the user interface for cluster deployment and testing tools is implemented, After the output of the performance data, the visual interface is generated to provide visual and visual cluster performance feedback.

Keywords: Container；cluster；Web；distributed；Kubernetes

目  录

摘  要I

AbstractII

第1章 绪论1

1.1课题研究的背景及意义1

1.1.1 提高资源利用率1

1.1.2 服务的动态扩容和持续集成1

1.2国内外研究现状2

1.2.1CasS平台2

1.2.2容器VS虚拟机——商用为主2

1.2.3微服务化3

1.3主要研究内容4

1.4论文组织结构4

第2章 总体设计6

2.1功能设计6

2.1.1Minion节点组件6

2.1.2 Master节点组件7

2.2基本概念8

2.2.1 Pod8

2.2.2 Service10

2.2.3 Replication Controller10

2.2.4 Label11

第3章 详细设计12

3.1 Master节点12

3.2 Kubectl12

3.3 API Server12

3.3.1 Minion Registry14

3.3.2 Pod Registry14

3.3.3 Service Registry14

3.3.4 Controller Registry15

3.3.5 Endpoints Registry15

3.3.6 Binding Registry16

3.4 Scheduler16

第4章 设计实现18

4.1 Pod18

4.1.1 Volume20

4.1.2 容器通信21

4.2 Replication Controller21

4.2.1 RestartPolicy和Replication Controller22

4.2.2 Replication Controller的工作机制24

4.2.3 Replication Controller的应用场景25

4.3 Service27

4.3.1 工作机制27

4.3.2 数据结构定义28

4.3.3 服务发现29

4.3.4 Headless Service30

4.3.5 入口地址和外部可路由性31

4.3.6 避免端口冲突31

4.3.7 Service入口32

4.3.8 存在的的不足33

4.4 API Server33

4.4.1设计目的34

4.4.2资源对象的RESTful接口34

4.4.3 API操作的原子性35

4.4.4 API Server流程36

第5章 调度和网络39

5.1 网络39

5.1.1 网络模型39

5.1.2 具体实现39

5.1.3 其他工作39

5.2 Minion管理41

5.2.1 Minion发现41

5.2.2 Minion管控流程42

5.3 调度器42

5.3.1 调度策略43

5.3.2 例1：Predicates.PodFitsPorts43

5.3.3 例2：Priorities.LeastRequestedPriority45

5.3.4 缓存模型45

第6章 实战部署47

6.1 环境准备47

6.1.1 下载软件包47

6.1.2 下载源码包进行编译47

6.2 upstart脚本47

6.3安装kubernetes客户端程序48

总结与展望49

参考文献50

致  谢51

附  录52