Dynamic reconfiguration of software systems can be achieved by exploiting variability models of such systems, combined with mechanisms for selecting and instantiating the appropriate system configuration. We developed a model-driven approach for dynamic software reconfiguration, which uses a component profile-oriented feature model for representing different configurations of a software system, and a grammar based genetic programming tool that, at run-time, automatically generates an optimal system configuration. The resulting feature configuration is transformed to a target format such as JSON, SQL or other specification that allows us to instantiate the new system configuration. In this paper we focus on the run-time reasoning and propagation aspect, and discuss our experience on applying the approach to a use case.