Some TSM client dedup file backups can be unrecoverable if server pool runs out of storage (IT11994)

Content

During IBM Tivoli Storage Manager (TSM) backup or archive operations using client-side deduplication, under certain conditions when the TSM server has little data storage space remaining, a file can be incorrectly backed up or archived by the TSM client with no error message for that file. The restore of an affected backup or archive copy will fail with an error such as 'digest validation error', 'unknown format', or a crash. The data that was intended to be backed up or archived in the copy being restored may not be recoverable.

APAR IT11994 has been opened for this issue.

RELEASES AFFECTED

The backup-archive clients in all TSM releases since client-side deduplication was introduced (in TSM 6.2) are affected, including the following levels:

• TSM 7.1.0 thru 7.1.3
• TSM 6.4.0 thru 6.4.3
• TSM 6.3.0 thru 6.3.2.4
• TSM 6.2.0 thru 6.2.5.4 (the 6.2 release has reached End of Support)

CONDITIONS REQUIRED

All of the following conditions are required for the problem to occur:

• Client-side deduplication was used for the backup or archive operation
• The backup or archive of a large file failed because the TSM server reported "out of data storage space" to the TSM client during the operation

Examples of circumstances that can cause the out of space error:

• Not enough space in the primary storage pool
• Not enough space in the storage container pool introduced with TSM (IBM Spectrum Protect) 7.1.3
• Not enough space in the next storage pool, if it is also a deduplicated storage pool
• Not enough scratch volumes
• Size of the large file exceeds the maximum object size ('maxsize' parameter) for the storage pool
• There is an additional file (or files) left in the backup or archive session after the large file that failed, and the very next file is small enough to fit in the remaining storage pool space.

DETERMINING IF YOU ARE POTENTIALLY AFFECTED

If you use client-side deduplication for your backup or archive operations, examine your client schedule and error logs. If you received these TSM client error messages when backing up or archiving a file: 'ANS1228E Sending of object filename failed' and 'ANS1311E Server out of data storage space', then if the operation completed on the next file without failure, the backup or archive copy of that next file could potentially be affected.

When an affected backup or archive copy is restored, the restore fails with an error such as 'digest validation error', 'unknown format', or a crash.

Use the script and query described in this technote to determine if your TSM server has affected files:

CIRCUMVENTIONS

Until you can apply a fixing level, use one of the following circumventions:

• use server-side deduplication instead of client-side deduplication
• turn off deduplication

FIX

The following client levels fix the issue for future backups and archives. There is no solution that correctly restores the backup or archive copy of an already-affected file. Contact IBM Support for assistance with possible recovery of data in an affected copy of a file.