This paper deals with real-time scheduling in a re-configurable multiprocessor system. Each processor is supplied with a renewable energy source, and uses a rechargeable energy storage. A reconfiguration scenario is any operation that consists in the addition, removal or update of tasks which may result in timing unfeasi-bility. In our work, each task is represented by a Proba-bilistic Directed Acyclic Graph (DAGP). Our contribution in this paper concerns the online feasibility problem that is issued from the reconfiguration process. We describe new approaches to solve this problem: the first one changes the timing parameters of the DAGP, the second one decomposes and migrates branches of the DAGP, the third one modifies the scheduling mode, and the fourth one deletes some tasks.