In the growing landscape of Virtualized Network Function (VNF) development processes and methodologies fueled by enabling technologies for virtualization, the myriad of customization options unveil unprecedented SW/HW configuration knobs and hazards. Underlying execution environments multiplex resources imposing hard-to-predict relationships between VNF performance metrics (e.g., latency), allocated infrastructure assets (e.g., vCPU), and stimuli workloads. Gym is a framework designed to enable automated testing and to extraction of such relationships by the means of VNF performance profiles, walking through a cause-effect path towards agile DevOps methodologies for NFV. The demo showcases the implementation of Gym through exemplified live extraction of VNF metrics for analytics use cases, such as comparison factors between physical network functions and pre-deployment infrastructure dimensioning.