Update 3/31/2011: The updated Exchange 2007 SP3 RU3 has been released. See Announcing the Re-release of Exchange 2007 Service Pack 3 Update Rollup 3 (V2).3/30/2011: We posted a status update for this issue. See Exchange 2007/2010 Rollup 3 Status Update.
Over the weekend, the Exchange Product Group was made aware of an issue which may lead to database corruption if you are running Exchange 2007 Service Pack 3 with Update Rollup 3 (Exchange 2007 SP3 RU3). Specifically, the issue was introduced in Exchange 2007 SP3 RU3 by a change in how the database is grown during transaction log replay when new data is written to the database file and there are no available free pages to be consumed.
This issue is of specific concern in two scenarios: 1) when transaction log replay is performed by the Replication Service as part of ensuring the passive database copy is up-to-date and/or 2) when a database is not cleanly shut down and recovery occurs.
While only a small number of customers have been affected to date, we believe the risk is significant enough that we are recommending all customers to uninstall Exchange 2007 SP3 RU3 on all Mailbox Servers and Transport servers. Uninstalling the rollup will revert the system back to the previously installed version. We have also removed the Exchange 2007 SP3 RU3 download from the Microsoft Download Center and from Microsoft Update until we are able to produce a new version of the rollup.
We are actively working this issue and based on test results plan to release an updated version of Exchange 2007 SP3 RU3 to the Download Center later this week. In addition, we are conducting an internal review of our processes to determine how to prevent issues such as this in the future.
When this issue occurs, the following similar events are logged in the Application Event log of the Mailbox server. Regardless of whether you see these types of events, you should review the recovery instructions and begin that process. If you are uncomfortable performing any of these steps please contact Microsoft Support for assistance.
In addition, in environments utilizing Continuous Replication, comparison of the database file between the active and passive nodes will indicate that the database file has decreased in size.
Regardless of whether you are experiencing this issue, we strongly recommend taking the below actions to ensure that you do not experience any data loss or outage event associated with this issue.
If you note the listed events in your environment the following steps must be taken in order to restore your high-availability configuration:
If you are not seeing these events in your continuous replication enabled environment, we recommend the following steps:
Kevin AllisonGM Exchange Customer Experience