Ever notice how when you move an Exchange Virtual Server from one node to another in an Exchange cluster, any other IIS websites you have created on the Exchange cluster are stopped during the move and not restarted?

The easiest way to experience this is to start up the “Default Website” in IIS Manager on the cluster node where the Exchange Virtual Server (and the “Exchange Virtual Server” HTTP virtual server) is running. Then, move the EVS from this node to another node in the cluster. After the move has completed, notice that the “Default Website” is now stopped on the other node. Start it up again in IIS Manager on this other node and move it back. Again, it’s stopped on the original node. You’ll also find this same behavior affects other HTTP virtual servers running in the cluster, so if you’ve defined other non-Exchange HTTP virtual servers to run in the cluster, only the Exchange HTTP virtual servers will be brought back online automatically after group moves or Exchange failures.

This happens because the ServerAutoStart bit in the metabase is reset to 0 for EVERY HTTP VS each time the Exchange resources move to another node in the cluster.

Don’t like this behavior? You’re in luck! There’s now a hotfix to prevent this from affecting other HTTP VS. With the hotfix in place, only Exchange HTTP virtual servers will have their ServerAutoStart bit manipulated. Have a look at KB.889171 for the Exchange 2000 hotfix or KB.885440 for the Exchange 2003 Post-SP1 hotfix.