Nurse Rostering Problems (NRPs) consist of generating rosters where required shifts are assigned to nurses over a scheduling period satisfying a number of constraints. Most NRPs in real world are NP-hard and are particularly challenging as a large set of different constraints and specific nurse preferences need to be satisfied. The aim of this paper is to show how NRPs can be easily modelled and efficiently solved using soft global constraints. Experiments on real-life problems and comparison with ad'hoc OR approaches are detailed.