Design And Implementation Of Consistency Anomalies Detection System In Microservice Transaction

Microservices is an emerging software architecture design principle that divides modules by services.A microservice transaction symbolizes a unit of work which span multiple services.Developers often rely on consistency anomalies detection tools to find inconsistent data in the concurrent execution of microservice transactions.In this paper,we propose a consistent anomalies detection system.The main work and innovations are as follows:1.Other systems always detect consistency anomalies from the perspective of data verification.In order to solve the problem of the data verification rules which are based on different specific businesses with high labor cost,we propose a consistent anomaly detection system based on Global serializability and Quasi serializability theory which are widely accepted as correctness criterions for the concurrent execution of distributed transactions from the perspective of serialization verification.By constructing a precedence diagram based on conflicts between transactions and finding all elementary circuits in the precedence diagram,we can find transactions with consistent exceptions without relying on any specific business.2.In order to solve the problem of current solution to obtain the conflict relationship which must modify source code and excessively invade the system,we design and implement a technical scheme based on data source agent.The scheme uses AOP programming ideas to obtain JDBC instances without modifying the source code.It can be quickly integrated by modifying the config files in the microservices developed by commonly used microservice development framework,and it is truly pluggable.In order to verify the effect of detecting consistency anomalies in this system,we implement test cases of consistency anomalies based on an e-commerce system composed of four microservices.Experimental results show that our consistency anomalies detection system can effectively identify execution histories which don't meet one of the global serializability and Quasi serializability standards.At the same time,the involved business and data items are sent to the microservice developers.