The 4th industrial revolution entails new levels of data driven value chain organization and management. In industrial environments, the optimization of whole production lines based on machine learning algorithms allow to generate huge business value. Still, one of the open challenges is how to process the collected data as close to the data sources as possible. To fill this gap, this paper presents an OpenTOSCA-based toolchain that is capable of automatically provisioning Apache Flink as a holistic analytics environment altogether with specialized machine learning algorithms. This stack can be deployed as close to the production line as possible to enable data driven optimization. Further, we demonstrate how the analytics stack can be modeled based on TOSCA to be automatically provisioned considering specific mock services to simulate machine metering in the development phase of the algorithms.