Greetings, Running Exchange 2003 sp2 in 2 node active /passive cluster 19 stores and around 900 mailboxes Occasionally we get event 9646 popping up in the app log on the exchange server where a user will have exceeded the maximum number of mapi connections and that user no longer can connect to their mailbox using Outlook until one of those mapi sessions times out. The most compelling suggestion I have found to resolve this was posted here: http://blogs.technet.com/b/jeff_stokes/archive/2008/11/25/how-to-get-rid-of-9646-events.aspx Specifically he recommends hardening the TCP/IP stack by changing the TCP KeepAliveTime to 5 minutes from the default of two hours as posted here: http://support.microsoft.com/default.aspx?scid=kb;EN-US;324270 Has anyone done this on an exchange cluster in a production environment? Did it get rid of your 9646 messages? Would you recommend it and why or why not? Thanks in advance

I had faced a similar issue with exchange 2003 sp2 . And it is very annoying. The issue is from networking side. I would normally update the nic rivers and switch firmware and educate the users about not openign many sessions ,like opening may folders and messages on outlook. Exchange 2003 is very stable and changing any settings may adversley affect the performance. The sessions are normally the Information store sessions and they are of different types . If these exceed 32 then the issue happens. But OWA works as it is http sessions has to cover another 32 . You can check for the settinsg under logons /tab in ESM. Right click Logins under server object and add colums . Here you can find different settings add those and verify if they cross 32 then the user starts complaining. There is a utility which can delete those sessions but it is bit dangerous as if you delete store sessions you may bring the IS down. I would check the uss affected by this problem from ESM and check the NICS and switches in the Path , upgrade their firmware , divers and work on it. You can use tcpview utilty with care to delete the sessions

On Wed, 23 Feb 2011 21:49:51 +0000, Nefarious-P wrote: > > >Greetings, > >Running Exchange 2003 sp2 in 2 node active /passive cluster 19 stores and around 900 mailboxes > >Occasionally we get event 9646 popping up in the app log on the exchange server where a user will have exceeded the maximum number of mapi connections and that user no longer can connect to their mailbox using Outlook until one of those mapi sessions times out. > >The most compelling suggestion I have found to resolve this was posted here: http://blogs.technet.com/b/jeff_stokes/archive/2008/11/25/how-to-get-rid-of-9646-events.aspx > >Specifically he recommends hardening the TCP/IP stack by changing the TCP KeepAliveTime to 5 minutes from the default of two hours as posted here: http://support.microsoft.com/default.aspx?scid=kb;EN-US;324270 > >Has anyone done this on an exchange cluster in a production environment? Yes. >Did it get rid of your 9646 messages? Some of them. Others were legitimate. >Would you recommend it Yes. >and why or why not? Dead connections from Outlook won't be used if the client reconnects. Why keep them hanging around? If you have users with mobile devices moving between WiFi access points the connections are dropped and new ones created just by moving around in the office. --- Rich Matheisen MCSE+I, Exchange MVP --- Rich Matheisen MCSE+I, Exchange MVP