As multi-source, component based platforms are becoming widespread both for constrained devices and cloud computing, the need for automatic logging framework is increas- ing. Indeed, components from untrusted and possibly competing vendors are being deployed to the same runtime environments. They are also being integrated, with some components from a vendor being exposed as a service to another one. This paper presents our investigations on an automated log-based architec- ture called LogOS, focused on service interactions monitoring. We ported and experimented it on Java / OSGi to enable identification between bundle providers in cases of failures. We motivate the need for an automatic logging framework in service- oriented architectures, and discuss the requirements of such frameworks design. We present our implementation on OSGi and expose the trade-offs in doing so. We conduct some experiments and, despite a necessary and significant existing overhead due to unequivocal identification constraints, we show that it should not be a major hindrance to the adoption of automatic frameworks for most service-oriented applications. Finally, we position our approach and give some perspectives.