MCIS SMTP Non-Delivery Report Occurs During Directory SQL Server Interruption

ID: Q221104

The information in this article applies to:

Microsoft Commercial Internet System version 2.0
Microsoft Site Server version 3.0

SYMPTOMS

When a SQL Server partition for a Site Server Personalization and Membership (P&M) directory fails over, messages may be returned with a non-delivery report (NDR). The NDR returned by the postmaster contains the following message:

Routing Table Database Transaction Failed

CAUSE

Default SMTP Delivery properties are set to 48 attempts, 60 minutes each, for both Local and Remote queues. This allows for a significant total period of attempts, and accommodates a relatively lengthy period of time during which P&M Authentication may be unavailable for a variety of reasons.

If the maximum number of retries has been dropped to two or three, this may result in a dramatically decreased period of effective delivery attempts. Only N-1 attempts are likely to occur if N maximum retries are configured. Due to additional processing delays, it is unlikely that the actual process of performing all N attempts will complete in the total time calculated. The final attempt that is expected is unlikely to occur.

RESOLUTION

To resolve this problem, increase the maximum number of attempts and the retry interval. This will help decrease the possibility of NDRs due to periods of LDAP P&M resolution failure. By increasing the number of retries, the resolution eventually completes, allowing the messages to be delivered when the SQL Server computer is returned to operation.

STATUS

Microsoft has confirmed this to be a problem in Commercial Internet System version 2.0, which is configured with Site Server Personalization and Membership version 3.0.

MORE INFORMATION

There are a number of reasons that the LDAP, P&M, or SQL Server services comprising directory authentication fail to respond for brief periods of time. This can be true even if redundant and fault tolerant systems are configured. Microsoft Cluster Server can be used to provide SQL Server fault tolerance. However, if a SQL Server computer providing storage for a P&M Store fails, the fail-over process takes a several minutes. Included in this time, is a period when P&M pauses for several minutes following the return of the SQL Server partition. The following sequence of events may result in NDRs:

Message Header Routing information look up is attempted by the SMTP Service.

The LDAP server cannot contact the SQL Server store temporarily due to a SQL Server system failure.

The Header information is not found through authentication and the "Routing Table Database Transaction Failed" error message is returned. The message is then placed in the local or remote retry queue.

LDAP returns to ordinary operation and new mail messages are delivered properly.

When the SMTP Retry Interval * (Maximum Retires-1) expires, the messages in the retry queue are handled as NDR. P&M Authentication must be operating properly before this period of time expires.

Note: A delay occurs following an event in which a SQL Server store becomes unavailable to the P&M LDAP services. This delay occurs after SQL Server has been returned to operation, and is in addition to the actual period of time that the SQL Server computer remains unavailable. Testing shows that it may take additional minutes for the services to begin to respond successfully again. Messages may fail P&M authentication and result in NDRs if the total period of P&M authentication response failure is greater than the total retry interval.

Additional query words:

Keywords :
Version : winnt:2.0,3.0
Platform : winnt
Issue type : kbprb