The Design And Implementation Of An Event-driven Operation Management And Service Registration System

In the search system of the Byte Dance,the distributed Search Retrieval service is a subsystem for retrieving the result of search.In order to ensure the stable running of this service and the correct discovery of its service instances by other related services,the developer and operations need a set of operation management system with low coupling with business logic to provide operation and maintenance support and service governance functions.Therefore,the operation management system needs to access the system to complete the operation and maintenance operations without invading the service code,and provide accurate meta information description for the business service.Given this,the Event is used to represent the notification issued by the system or operations when the state of the business system changes in the thesis.And the the-sis proposes to use Event-Driven architecture to design the system architecture.The different parameters and their values carried by various events,which represent the objectives and specific operations involved in the corresponding operation tasks,in or-der to perform various types of operation and maintenance operations required by the user.The communication between operation systems and business systems through event-carrying commands and parameters can reduce the coupling of offering opera-tion support for the business systems,and reduce the cost of business accessing op-eration system.At the same time,considering the entities involved in the running of the business,including clusters,databases,and service instances,the system maintains metadata that accurately describes the current state information of the service instance,and help the business system complete service registration.The main body of the system is developed with Python version 2.7.Distributed MySQL database is used to storage the data entity status information maintained by the system and the event records circulating in the system.And the system uses the Flask framework to develop the HTTP API of the system.As well as,it use the gflags command-line parameter parsing tool to provide command-line language user inter-face,using Consul client to register services,and accessing the distributed database.Distributed lock based on ZooKeeper is used to ensure data consistency in the mean-while.Each module of the system is deployed in an isolated node based on Docker container in the form of a separate process.The thesis completes the design and implementation of the overall architecture and main functions of the above system.After the system version 1.0is launched,it provides operation support for the company's search retrieval service,and is respon-sible for maintaining the life cycle of service clusters and databases,and performing registering of service instances' information.It reduces the workload and misopera-tion probability of manual operation,and facilitates the operation and maintenance of thousands of servers in the cluster,providing support for the stable operation of the business system.