We are migrating our users from our Exchange 2003 servers to our new clustered Exchange 2010SP1-servers. We have already converted 20.000 student accounts and around 2.000 staff members. We have 2 CAS servers in a NLB-cluster and 5 Mailbox servers in a DAG-group + 2 dedicated Public Folder-servers. Until now everything was working fine, but starting last weekend, some outlook-clients began losing their connections (or cannot even make a connection) to our Exchange servers. 95% of the clients is not affected, but the remaining 5% is a mixture of 2003, 2007 and 2010 Outlook-clients. OWA on these workstations works fine, and we tested with or without virusscanners and local firewalls.
We looked at the workstations, but their configuration was ok: some were configured in direct mode, some using cached mailboxes. The effect are problems starting up, or even freezes when working for a few minutes: they can open their list of mails, but when they try to open a mail, Outlook hangs. The load on the CAS servers is between 5-15%, with peaks at 30%.
If we switch the communication protocol on these clients to https instead of TCP/IP (Outlook Anywhere or RPCoverHTTPS), they work fine.
The only changes we made recently is install rollup 2 for 2010sp1 on both cas-servers. Is there a known issue that this will influence the TCP/IP-communication between clients and server?
We are also moving more and more users each night, which results in an increased usage of the databases and client communications. I found an article about the client throttling in Exchange 2010 SP1 and found out that our RCAMaxConcurrancy was still al the RTM value of 20, but had to be changed to the default SP1 value of 214748364. I changed it, but there seems to be no improvement.
so, now my questions:
1) Are there known problems with rollup2 for Exchange 2010SP1?
2) Can I safely remove this from our CAS-servers to check if this is generating the problem?
3) Which settings can be/has to be changed in the Client Throttling to improve Outlook TCP/IP communication?
4) Can those throtlling-parameters be changed on the fly, without a service- or server-restart, and are they changed instanteaneously?
5) Is there something else that can cause this strange behaviour on -some- Outlook clients?