A buffer sizing method considering the activity duration risk as well as the multi-resource constraint risk under uncertainty is proposed. First of all, the key risk factors are identified and the activity duration risk is effectively assessed by using Bayesian networks. Then, the resource constraint risk is appropriately determined by using the resource flow networks. And a reasonable buffer sizing approach is developed for the construction of a stable critical chain scheduling plan. A numerical example is given to illustrate the effectiveness and practicability of the proposed method which is capable of maintaining the schedule stability with a short project makespan, while ensuring a high on-time completion rate.