This paper proposes a dynamic QoS assessment approach for Internetware with uncertainty reasoning. The salient feature of this approach is that it's based on Bayesian network, and can correctly predict the component's service capability in various combinations of users' QoS requirements, especially to the entities with different service levels. Through online learning, it supports to update the Bayesian network dynamically and improves the accuracy of prediction result. The proposed QoS assessment model has been implemented in the developed Service Coordination Bus (Once-ESB). The experimental results show that it is effective.