Tape drive problem with make_tape_recovery

[Brian Willis]

I have a HP K200 series 800 running HP-UX 10.20. (HP-UX hostname B.10.20 E 9000/819) It has 2 externel drive chasis and an internel tape and cdrom drive. I'm having some serious problems with the tape drive. I try and run a make_tape_recovery and get this error:
897+0 records in
896+1 records out
I/O error
ERROR: /usr/bin/dd if=/var/tmp/uxinstlf.recovery of=/dev/rmt/0mn bs=2k failed
to write LIF onto tape.
ERROR: Failed to generate LIF on tape .


======= 04/05/06 15:39:21 GMT make_tape_recovery completed unsuccessfully

Also this shows up in dmesg:
SCSI TAPE: dev = 0xcd030040 I/O error during close

This leads me to believe the tape drive is bad. A reboot does not fix the issue.

So 4 drive replacements, 1 daughter board, 1 system board, and a complete hardware swap we're having the exact same issue. The only time it's fixed is when the drive is replaced. For several weeks after a drive replacement it works fine. Then all of a sudden it starts acting up again. The drives have all been from HP and have been the same drive. The thing that strikes me as odd, is the dd output is exactly the same every time it fails. It's always at 896+1 on the write that it fails.

The only things that are the same on this "new" machine are the externel disks (which are on a different scsi bus) and the OS install.

Can anyone think of the root cause other than maybe scsi termination? Is there any good way to troubleshoot this other than waiting for it to fail?

This machine must stay in production through August of 2007 as there are still some in the field so I really can't imagine not having recovery tapes until then. Aside from being slow, the machine works fine.

[Steven E. Protter]

Shalom,

i think you are right. Its either the tape or the drive.

dmesg -
wait a bit
dmesg

You may see something.

SEP

[Brian Willis]

The only thing I see in dmesg related is the
SCSI TAPE: dev = 0xcd030040 I/O error during close

I've tried many different tapes and the tapes work in other machines. This drive was working fine until in was made into this new machine. Seems odd that regardless of the hardware, it eats entire tape drives.

[DCE]

Brian,

Have you gone to the patch depot and searched for SCSI patches, or for DDS patches? The 10.02 patches are still available,and there may be one or two that possibly address your problem.

[Brian Willis]

I haven't searched for any particular patches, but we run regular patches. We just run the patch tool and install all relevent patches. Would this have missed any?

This almost seems like a software problem that causes a hardware problem.

[John Dvorchak]

It has been my experience that when you get the LIF error the tape or the tape drive is bad. Since you have tried the tape in other machines and it works, I would look at the tape.

The easiest way is to try to tar something to the tape and see if that works:

The first thing I would do is to run a cleaning tape through the drive a couple of times then:


tar -cvf /dev/rmt/0m /tmp/somefile

then look for tar errors if none then:

tar -tvf /dev/rmt/0m

to see if it reads from the tape. I'm betting that the tape drive, especially if it's a couple of years old, just broke.

[John Dvorchak]

You didn't mention, or I just missed it, which drive this is. Is it a DDS-1 or DDS-2 drive? The older tape drives are prone to giving up the ghost. Since those drives are no longer in production the best HP can do is replace a failed tape drive with a drive that had failed earlier, possibly on another customer's site, tried to test/rebuild it and having no errors in the lab, re-deployed it to you.

If you have a hardware contract with HP you might suggest that they replace it this time with a new or newer DDS-3 drive. I think since this has a history, they might be able to get that approved.

[JASH_2]

Brian,

I realise that it is much more straight forward to recover from a tape, if needs be, but have you considered using one of your other machines as an ignite server and send a make_net_recovery image to one of them instead of using tape?

Just a thought.

Regards,

JASH

[Brian Willis]

Taring up files does return errors. The tapes work fine in other machines so it has alway led me to believe that the drive is bad. Probably a good point that all the ones I'm getting could just be going bad.

It is a DDS-2 drive. That's a good idea to change for a DDS-3 drive.

Also, the make_net_recovery sounds like a good idea, but I've never done it. Any good documentation for it? How would you boot the machine to get it to restore from the other server?