The problem of exception notification mechanism of Web services (WS) is identified in this paper, which has not been noticed or discussed before. The problem is especially critical since Web services are frequently upgraded due to its shorter lifecycle, e.g., from weeks to months, compared with traditional software development. Each time a service is upgraded, it should first be stopped, then upgraded, and finally restarted This will definitely affect the working schedule of clients. Some may receive time-out message and then stop working. Some may receive wrong results or even worse. Hence, this problem should be solved before WS clients can experience better service quality. Two preliminary solutions to the problem based on watchdog monitoring systems are proposed and their advantages/disadvantages are also discussed.