XADM: Unable to Back Up Exchange Server 5.5 with Event ID 105 (188646)

SYMPTOMS

When you are trying to perform an online backup of the Microsoft Exchange Server databases, the operation may fail and the application event log may display events similar to the following:

   Event ID 105
   Source:  ESE97
   Type:  Error
   Category:  Logging/Recovery
   MSExchangeIS ((215) ) The database engine has stopped the backup with
   error -1018.

   Event ID 118
   Source:  ESE97
   Type:  Error
   Category:  Database Page Cache
   MSExchangeIS ((215) ) Direct read found corrupted page (826096) with
   error -1018. Please restore the databases from a previous backup.

The event viewer may also display events similar to the following during online defragmentation:

   Event ID 184
   Source:  ESE97
   Type:  Error
   Category:  Online Defragmentation
   MSExchangeIS ((215) ) Online defragmentation of database
   'D:\exchsrvr\MDBDATA\PRIV.EDB' terminated prematurely after encountering
   unexpected error -1018.

   Event ID 116
   Source:  ESE97
   Type:  Error
   Category:  Database Page Cache
   MSExchangeIS ((215) ) Synchronous read page checksum error -1018
   ((1:826096 1:2184256518)
   (4119191402-546568656) (0-24859366)) occurred. Please restore the
   databases from a previous backup.

CAUSE

In certain database operations such as, but not limited to, online backup, the backup routine makes a call to the operating system to read a 4-KB page of data from the database on disk and write it to tape. Before committing the data returned from the operating system call to the tape, the online backup process compares the checksum value in the page header (recorded when this page was written to disk) to that being returned from the READ call. If the checksum values do not match, the Exchange database engine detects this and returns -1018 (JET_errReadVerifyFailure).

WORKAROUND

First determine the root cause of the corruption; that is, check the system log, run chkdsk, and so on. If the cause of the -1018 is not determined and fixed, the corruption problem will recur.

In order to perform a successful backup, use one of the following methods:

Restore the information store from the previous backup.

NOTE: If the information store cannot be started after restoring from backup because a "corrupted" log file is encountered, the log files may need to be removed. This will cause loss of data since the time of the last backup.
Move Users to another server. Moving users will not also move the corruption, because the checksum errors also cause the move process to fail for the affected mailboxes. Any mailboxes that fail the Move Users process should have their contents downloaded to PSTs.

NOTE: The specific messages that are directly associated with the corruption may fail to move to a PST and would then be lost.

The "corrupted" but now empty Priv.edb file may be deleted from the original server after users are moved off. The information store, when re- started, causes a new Priv.edb file to be generated. After this is done, the users may be moved back to the original server. The root cause should be determined prior to moving users back to the original server to prevent recurrence of any corruption problems.

MORE INFORMATION

The event viewer system log may report array errors similar to the following:

   Event ID:    9
   Source:      aic78xx
   Description: The device, \Device\ScsiPort0, did not respond within the
                timeout period.

In almost all cases, these messages are being posted due to hardware problems with either the controller or, more likely, a device that is attached to the controller in question. The hardware problems can be associated with poor cabling, incorrect termination or transfer rate settings, lazy or slow device responses to relinquish the SCSI bus, a faulty device or, in very rare cases, a poorly written device driver. For more information, click the following article number to view the article in the Microsoft Knowledge Base:

154690 How to troubleshoot Event ID 9, Event ID 11, and Event ID 15 error messages