Exchange administrators should be familiar, if only in passing, with the nuts-and-bolts details of TCP/IP. Many...
problems which appear to be a malfunction of the Exchange server are in fact due to network conditions. These can be difficult to troubleshoot without some understanding of the way TCP/IP works.
One of the parameters used by TCP/IP is the MTU, or Maximum Transmission Unit specification. This is a number that a computer's TCP/IP stack will use as the maximum size of a TCP/IP packet. Windows usually uses 1500 as the default MTU setting, but if TCP/IP is being encapsulated through a WAN or sent through an "intelligent" router or firewall that modifies the MTU, the results may confuse an Exchange server. A smaller message (one which fits into a frame of 1500 bytes) may get through, but a larger message may get dropped.
A common symptom of this problem in an Exchange server can be seen if the server's conversations are logged -- the conversation goes through the HELO, MAIL FROM and RCPT TO stages, but "hangs" or times out at the DATA portion of the conversation. Often the symptom is all the more confusing when someone tries to reproduce the problem by trying to telnet into Exchange server's IMS and generating a message manually, only to have that work. (The reason a telnet conversation may work while others don't may have to do with the fact that the telnet conversation is not following the same route as a real message and therefore is not having its MTU modified -- or because the shorter packets in a telnet conversation never trigger the problem.)
Troubleshooting this problem often involves following each step the packet takes from the Exchange server to the outside world. Smart routers, VPNs, firewalls, and especially smart SMTP relay hosts should be considered as the main culprits for this condition.
Serdar Yegulalp is the editor of the Windows 2000 Power Users Newsletter.