Detecting Interaction Related Bugs in a Multi-Tenant Setting Using Kubernetes

Multi-tenant architectures allow multiple users to run similar applications on shared infrastructure, reducing the cost of resources. Developers of such applications must ensure tenant isolation to prevent users from interfering with one another. Concurrency bugs may arise only when a specific set of users is logged in at the same time. This paper introduces a framework for testing user interactions with a server, aimed at detecting timing-related bugs. The framework uses a scheduler to control when users log in and out of the system. If a particular set of users is logged in simultaneously, the server will simulate a potential bug. Users also keep a copy of the server’s data, and if the server’s data differs from their copy, a bug has occurred, which the user will report. The framework is designed to be flexible, supporting the addition of custom schedulers that can use either static scheduling algorithms or advanced Artificial Intelligence techniques to speed up bug detection more efficiently than traditional fuzz testing methods.