The exponential traffic growth and the increasing complexity of WDM networks demand a more sophisticated control protocol that will provide a more efficient use of optical network capacity. Many researchers propose to extend GMPLS protocol suite to activate WDM networks, such as by enhancing OSPF-TE, RSVP-TE, etc., the decentralization of control functions is also an inevitable trend of developing WDM networks, when the network continues to scale up. However, the control difficulty is raised by the lack of accurate network state information in this distributed system. Therefore, an effective measure is required to tolerate a certain inaccuracy of such information without significantly increasing the overhead of the control system, such as the traffic volume of the control packets, connection set-up delay, etc. In this paper, a dynamic routing and wavelength assignment with multi-wavelength probing is conceived, which extends our previous case-study toward translucent wavelength convertible WDM networks. For the simulation purpose, we assume that wavelength converters have been sparsely placed in the network, and inside these nodes, the optoelectronic 3R regeneration is potentially obtainable too. The path computation element is served as the calculation unit for the control tasks. In experiment, not only the blocking performance of the proposal is investigated by comparing with traditional single or full probing methods and with centralized control case, but also the control overhead is measured, e.g., update traffic volume, set-up delay. The results reveal that the multi-wavelength probing approach trades off the above two aspects, and is proved to be well-suited for distributed WDM networks.