Tape wearing out on Exabyte?

[Jeff Mittelman]

This is not normal.  This is a known bug in the SCSI driver.  The bug
number is 1042822 and a patch is supposed to be available in the next
couple of weeks.

Exabytes normally get a certain number of retries when writing or reading
a tape.  If you want to know how many have occurred on the current tape at
any point, do a "mt status" on the drive.

The number of retries you get depends on the tape, how clean the heads
are, and how much of the tape you have used so far.  When writing
full-length 2 Gb tapes with an apparently-clean drive using new Sony
tapes, I've seen anywhere from 200 to 20,000 retries.  I can only assume
that this is mostly variation in the quality of the coating on the
individual tapes.

To complicate things further, the Sun SCSI tape driver seems to report the
number of retries only when the device is closed, *and* only if the number
is greater than about 5000.  So you can go along for years without ever
seeing that message if you typically write only a few hundred Mb to a tape
(e.g. for backups).  Then, when you start writing full-length tapes, a
certain percentage of your tapes (10-20% of mine do this) get enough
retries for the errors to get reported.

In general, you don't need to worry about it.  The drive *has* rewritten
those data blocks and should be able to re-read the tape without problem.
If it could not write the data successfully, it should have given you an
I/O error.  It does take extra space on the tape to do these rewrites, but
the drive does not normally use the full length of the tape anyway, so
there is built-in capacity for a certain error rate.

You might want to try cleaning the drive when you see the message, just in
case the head is dirty.  But in my experience, the errors seem to be
mostly associated with a particular tape.  I've had one tape get over
10000 errors, then the next tape in the same drive would get only 2000,
without doing anything at all to clean the drive.