Summary: This paper analyzes possible anomalies in Web Services execution during a composition, it introduces different strategies to face these anomalies and a solution for fault tolerance on Web Services compositions based on synchronization. This solution is as follows: when an anomaly is found, the process calculates the affected area and it rolls back the execution to the closest synchronization point that covers it, then it uses services substitution or partial replanning to fix the problem and to go on with the execution of the composite process.