Modeling and Automated Execution of Application Deployment Tests

Abstract

In recent years, many deployment systems have been developed that process deployment models to automatically provision applications. The main objective of these systems is to shorten delivery times and to ensure a proper execution of the deployment process. However, these systems mainly focus on the correct technical execution of the deployment, but do not check whether the deployed application is working properly. Especially in DevOps scenarios where applications are modified frequently, this can quickly lead to broken deployments, for example, if a wrong component version was specified in the deployment model that has not been adapted to a new database schema. Ironically, even hardly noticeable errors in deployment models quickly result in technically successful deployments, which do not work at all. In this paper, we tackle these issues. We present a modeling concept that enables developers to define deployment tests directly along with the deployment model. These tests are then automatically run by a runtime after deployment to verify that the application is working properly. To validate the technical feasibility of the approach, we applied the concept to TOSCA and extended an existing open source TOSCA runtime.