Hi,

we had a planned power outage recently, and during bringing everything
back to service the harddisk of one Linux client failed. Restore to a
new harddisk completed in reasonable time, but unfortunately some files
had bit errors, rendering some data unreadable (there is an encrypted
filesystem running on this client).

The drive that was used reading this tape was replaced by IBM service
very shortly afterwards because of read errors (ANR8944 in TSM Actlog),
but even restoring the whole data from the vault had no effect, the bit
errors in the data remained.

To make things worse, there was a TSM client 4.1.1.0 running on this box
that was at the beginning running a Suse 6.3 OS (now nearly every
software self-compiled), so I don't think I'll be getting help from TSM
support here. The first restore was made using a Knoppix CD with TSM
client 5.4.1.2, the second using the original TSM client (4.1.1.0).

I was told by the client's administrator that only newer files were
affected by these bit errors. I haven't changed anything at TSM server
levels recently, but upgraded AIX from 5.3 TL5 to TL6 and the firmware
of the pSeries to the latest level.

Are there any known issues that may have caused these bit errors in the
data, especially in the combination very old Linux client and (quite)
new server? Is it possible that the AIX and/or firmware upgrades have
had such an unwanted side-effect?

Any help or tips are appreciated.

TSM Server 5.3.4.2 running on AIX 5.3 TL6
3494 tape library with 3592 Drives (J1A in the logical partition in
question)

Regards,

--
Bjoern Rackoll
University of Hamburg
Regionales Rechenzentrum
Zentrale Dienste
Schlueterstr. 70
20146 Hamburg
Tel.: +49 (0)40 42838 - 63 11
Fax: +49 (0)40 42838 - 62 70
E-Mail: backup AT rrz.uni-hamburg DOT de