Design And Implementation Of Distributed Configuration Center Based On Raft Protocol And RocksDB

In the process of system development,some parameters and variables that need to be changed are generally separated from the code and managed separately as independent configuration files.As the function of the program becomes more and more complicated,the configurable items of the program are also increasing.Traditional configuration center solutions such as Taobao Diamond,Baidu Disconf,and Qihoo Qconf have shortcomings in scalability.At the same time,due to the lack of support for data consistency in the architecture,only the final consistency is met.So it is unable to provide a valid configuration management service for distributed programs.This paper intends to solve the above problems by combining the Raft consistency protocol with RocksDB.First of all,this thesis has carried out in-depth research on distributed system related theory,distributed data consistency algorithm,high availability and high performance server cluster architecture.After comparison,this paper chooses the Raft consistency protocol which is more suitable for engineering projects and the excellent performance RocksDB to implement the distributed storage framework.Secondly,through the analysis of the relevant configuration management use cases in reality,the functional requirements of the basic configuration requirements,advanced management function requirements and client requirements of the unified configuration center are proposed.In addition to this,the paper also presents non-functional requirements such as data consistency,stability and high availability,scalability and security.Then,analyze the shortcomings of the existing configuration center in the architecture,real-time push mode and grayscale release mode.This paper proposes how to realize the distributed storage model based on Raft coherence protocol,how to realize the real-time push of configuration and how to realize the three key technical problems of grayscale publishing and historical configuration rollback,and give corresponding solutions in turn.Finally,this paper designs and implements a distributed configuration center with real-time push,grayscale release,and configuration rollback functions,and performs deep functional testing and performance testing.