This Knowledgebase article provides information about how to perform a switchback after performing a switchover in an Exchange environment.
Every domain controller in the domain runs the Kerberos service and is capable of granting session tickets for users and computers. If a domain controller is not available, domain authentication fails and the user is logged on by cached logon credentials at the client computer with no secure channel to the domain.
The client computer periodically attempts to locate the Kerberos service every fifteen minutes or when other processes attempt to connect to domain resources during the user's session, and will complete the domain authentication process if one is found.
On a switchover, the secure channel may not be established in a timely fashion before the system attendant that relies on the secure channel begins to start. The time to establish this connection may vary depending upon several environmental factors, most notably the load on the domain controller and the topology of the domain.
This is not normally an issue with a single switchover unless the network connectivity to the domain controller is slow or the domain controller is busy, however if a switch is followed by a switchback within a 10 minute period, Microsoft System Attendant may fail to start because the secure channel has not been re-established before the system attendant begins to start.
After performing a switchover, Neverfail recommends waiting at least 10 minutes before performing a switchback. Otherwise, Microsoft System Attendant may fail to start resulting in the following errors:
Error General 1005 Unexpected error The operation failed. ID no: 80004005 Microsoft Exchange System Attendant occurred.
Error General 9004 The Metabase Update service failed to start, error '80004005'.
Error General 1002 Metabase Update agent failed to start. Error code is 80004005 (Unspecified error).
For more information, please see:
To ensure system stability and guarantee application start-up, Neverfail strongly recommends that a switchback is not performed within 10 minutes of a switchover or failover occurring.
If the symptoms were observed without switching quickly (under 10 minutes) then the topology may be to blame and increasing the discovery time for the system attendant to find the domain controller should be adjusted as per Microsoft Knowledge Base article #312859 .
If this still does not cure the symptoms the secure channel can be forced to establish prior to the system attendant starting stopping and starting the netlogon service. This can be achieved by including the following at the start of the application start scripts.
Net Stop Netlogon
Net Start Netlogon
Should System Attendant fail to start in the circumstances described then the system administrator should proceed to the 'Start/Stop' tab in 'Application Manager' panel in the Neverfail Heartbeat Management Client and press Start .
Tip: Neverfail Heartbeat should be configured with the appropriate alerts ('net send' or another mechanism) to warn system administrators that an application failed to start.
This applies to SP 2, 3, and SP 3.1 with Exchange 2000 and 2003.
Note: If you continue to experience problems with System Attendant after following the above advice, please contact your Neverfail Support representative.