Recovering Gropius Models with the Cluster Architecture Recovery Assistant

Modern software systems often follow a component-based architecture style, e.g., microservices. Despite many advantages of such an architectural style, cross-component issue management is a complex and time-consuming challenge. The Gropius issue management system enables cross-component issue management with respect to the issues' architectural dependencies but requires an architecture model of the system. While creating such a model for new projects is straightforward, it is often unfeasible for older projects due to the large modeling effort required. One solution to this challenge is to recover the system's architecture and map it to a Gropius model. However, as different projects can use different programming languages, the recovery must be as programming language-independent as possible. Therefore, in this work, we present the Cluster Architecture Recovery Assistant (CLARA), a framework that combines various established dynamic architecture recovery approaches and a new approach based on DNS scraping. CLARA allows recovering of a component-based architecture that is deployed on Kubernetes and maps this architecture to a Gropius model. We evaluated CLARA by applying it to an open-source microservice reference architecture and a microservice application of Volkswagen AG, and we compared the recovered architectures with the actual ones. The results indicate that CLARA can recover Gropius models effectively but show some limitations, e.g., detecting fine-grained relation types.