Media error on Exabyte

[Dave Olson]

In <muenzel.657291850 at wega> muenzel at wega.tat.physik.uni-tuebingen.de (Stefan Muenzel) writes:
| Once more a problem with the EXA-Byte:
| We have a EXA-Byte 8200 and want to use it with a SGI 4D/20.
| But when we try to tar some bytes we get messages like the following:
| 
| Oct 22 18:05:04 mira unix: NOTICE: SCSI tape #5 Uncorrectable media error
| Oct 22 18:05:04 mira unix: NOTICE: SCSI tape #5 Unrecoverable error
| Oct 22 18:05:04 mira unix: NOTICE: SCSI tape #5 had 2215 recoverable Errors
| 
| (Copied from the SYSLOG-file).
| This occures after some 100 kbytes. Our first guess was, that we had 
| bad tapes, but we've tried several new ones from Sony and Exabyte. It
| seems not to be a problem of the SCSI-bus.

The message indicates that the Exabyte returned a check-condition on
a write, then returned 'media error' as the specific error type.

The recoverable error message indicates that there were a LOT of errors
that were recovered from by retries.  If all this happens in 100 Kbytes,
then I see several possiblities:

1) your heads need cleaning.  Get the dry (not wet) cleaning kit and
   clean the heads.  This can make a big difference.  Remember that 
   the heads should be cleaned on a regular basis, and more often
   when using new tapes, since they tend to shed more coating.

2) Get certified data tapes if 1) doesn't work.  In any case, we
   have been told by Exabyte tech people to avoid Maxell and TDK,
   (as of a year OK; the situation may have changed since then)
   and to NOT use MP (metal particle) tapes.  Sony tapes seem to
   work pretty well (which isn't surprising, since they make the
   Exabyte Data tapes).  A number of other brands seem to work OK
   also.

3) Watch your recovered error rate.  I typically see a few hundred
   for multi-Mb archives.  Whatever your rates are, keep an eye on
   them, and clean the heads when the error rate rises, or if you
   are trying new tapes, don't use that brand/type if they consistently
   return higher error rates.

   The recovered errors are reported on close of the tape if the
   tape was read or written (or other commands that moved the tape
   were used).  Typically the counter is reset only when the tape
   is unloaded.

--

	Dave Olson

Life would be so much easier if we could just look at the source code.