SCSI Tape Error

[Rob Smith]

Hello All,

One of my backups failed last night and this error appeared in both dmesg and syslog:

SCSI TAPE: dev = 0xcd020000 I/O error during close

I ran diagnostics on the drive and everything checked out okay. Has anybody ever seen this before? Any help would be appreciated. Many thanks in advance.

Rob

[A. Clay Stephenson]

Hi Rob:

The only thing that error message tells me is that it was a tape drive [major device 205 (0xCD hex)] and that it was c2t0d0. You may have more diagnostics in syslog - look for LBOLTS.

Having said that, tape drives occasionally fail
and will then work flawlessly for months. Remember there are a lot of mechanical parts that have to work flawlessly including the media itself. If robotics are involved even more things have to work. I always keep a log of any hardware incidents and this may help you to spot any trends among seemingly unrelated events. Since your drive passed it's diagnostics I would next run a few tars or cpios on the drive to make sure that not only is the tape drive functioning but also your connections to it. If it passes those tests, I suspect you are fine.

Clay

[linuxfan]

Hi Robin,

Like Clay mentioned, i have seen cases a tape drive would complain but then run without any problems. In the past, i have seen cleaning the tape drive helps. Also not sure if you looked at cstm/xstm to see if you see any errors reported there. just a thought.

-Regards
I am RU

[Dan Wanek]

You may also want to check how many devices you have on the bus. We've had problems similar to this when we had a lot of SCSI traffic on one bus. (i.e. more that 2 DLT drives)

[Bill Hassell]

This was probably caused by using the no-rewind device file without the Berkley option. AT&T has a (somewhat bizarre) behavior when closing with no rewind. You might expect the tape to stay in place...nope. Check the man 7 mt page (and also man 1 mt). You'll see that Berkley behavior is intuitive and should be used.

Here's what happens with AT&T: When a no-rewind AT&T device file is closed, the tape is moved to the beginning of the next file! If we were in the middle of the last file on tape, it will try to position to a non-existant position past the end-of-data.

So, if you want intuitive behavior, use the device files with Berkley behavior set. As always, never use ll to decode a device file, use lssf as in:

lssf /dev/rmt/*

[A. Clay Stephenson]

Hi Bill:

I don't think so in this case as the minor device number indicates he was using a rewind device. Otherwise dead on, always use the Berkeley devices. I always use the mnb devices in OB2 and very rarely I get this same sort of error in no repeatable pattern. It may not repeat on the same tape drive for months.

Regards, Clay