I have Unity 3.1.5 running in failover. Exchange 2000 is running on a cluster for redundancy. Both exchange nodes are domain controllers and global catalog servers. When Node A goes off line unity is no longer able to deliver messages. According to TAC they think it is due to global catalog being offline. The following registry keys are pointing to node A:
If the GC goes offline, there is a chance that messages may not be delivered. That may or may not happen depending the configuration. The API that Unity uses to send and receive messages (MAPI) does rely on a GC when there is an Exchange 2K environment. The GC that is being used by MAPI is actually specified in the particular profile that's being used for MAPI, not in these registry keys. And the GC that MAPI is using may or may not be the same as Unity's directory monitors (the directory monitors, the services that use the settings you posted, have nothing to do with message sending/receiving).
When you mention that messages are not being delivered, does that mean that Unity is giving errors, or everything appears to be OK, but the message never shows up? Has message tracking been enabled on the Exchange server?
Cisco is working with Microsoft on the issue of MAPI having problems when the GC goes down. Since the MAPI implementation is not written by Cisco, there's no direct coding that we can do to work around any MAPI/GC issues.a
Having said that, the directory monitors certainly do not handle a DC or GC going down well at all. That is something that we are currently addressing and you can expect to see improvements in that area around the 4.0.3 timeframe.
Getting errors is much better than the submission of the message saying it was successful, when indeed it was not. This is the way that I was expecting (and hoping, actually) that this would fail.
That error message means that there were no recipients on the message that Unity was trying to submit. This makes total sense if you think about how the MAPI profile uses the GC. It uses it for name resolution. It's kinda similar to when you type in a name in the "to" field in Outlook and the name is resolved to a mailbox. Since the GC is down, that name resolution doesn't take place, and MAPI can't resolve the names Unity has entered for recipients.
This is the issue that Cisco has engaged Microsoft on. Currently, the only workaround would be to manaully change the GC that the MAPI profile points to and restart Unity. If MAPI needs to resolve a name, and its server is unavailable, this problem will continue.
To find the exact profile is kinda tricky, as it will be in a different location depending upon the account that the AvCsMgr is running as. The easiest thing to do is open Regedit and search for "Unity System Profile". There will be an entry in the Unity (Active Voice, really) section of the Registr, you can ignore that one.
The ones that we'll be interested in are HKU\\Software\Microsoft\Windows NT\CurrentVersion\Windows Messaging Subsystem\Profiles\Unity System Profile
Look for the key under that starts with "DCA". In there you'll see a value for a servername. That's the GC. Let me know if that works out.
Thanks . If I restart the Unity services it works. My main concern is that the solution my customer bought is supposed to be fully redundant. And the problem becomes a single point of failure. I see bug CSCdu56529 is assigned, hopefully this gets resolved soon, or it will be hard to tell customers unity can be fully redundant.
Thanks for all your help. As usuall you guys have been a great help.
Are you getting this error “Installer User Interface Mode Not Supported. The installer cannot run in this UI mode. To specify the interface mode, use the -i command-line option, followed by the UI mode identifier. The value UI mode identifiers...
The below trick might come handy when you have to add a new node to a cluster but you don't have or is unsure of the security password for the publisher. This procedure has been around for ages.
1) Login into the CLI of the Publisher.