tape drive is unavailable

[mustafa_12]

Hi everyone,

When i tried to copy a file from disk to tape drive, the process hang and it did not return command prompt for about 8 hours. So, i cancelled the process (ctl-c). But when i look at the device with "show dev mga1" now, the output says that device is used by another process. Now, i can not use the tape drive and our backup jobs are waiting. When i search the case on internet, it is said that the only solution is to reboot the server. But it is a production side, it is not so easy to reboot, as you know.
Anyway, i can reboot the server by any means. But I wonder the reason and fair of repetition of the same condition. In attached file, i am sending some useful outputs. In product history, it is seen that there is no fibre_scsi patch (v1 is removed and there is no fibre_scsi patch. this is done by local HP site). May the reason be that?

Any comment will be appreciated..
Thanks in advance

[Aaron Lewis_1]

mustafa, a work around I've done in the past is to change the SCSI id on the tape drive and run IO AUTO from SYSMAN. This will give you access to the drive, and you can reboot later.

[Vladimir Fabecic]

AS I can see, tape drive is FC (mga1:). If you change SCSI ID you will have to do:
$ MCR SYSMAN IO FIND_WWID
and then
$ MCR SYSMAN IO AUTO
It may but also may not help.
First try to kill proces that is "holding" tape.
$ SHOW DEV MGA/full
Kill the owner process with STOP/ID=
Also try to logout and login after some time.
Tape devices have allways been problem maker in VMS and I remember situations when anly reboot helped.

[mustafa_12]

Aaron and Vladamir, thank you very much for your fast reply and suggestions.

I have changed the scsi id of the library.
("Menu -> Edit Options -> SCSI -> Drive 1 Bus ID" was 1. It was changed to 2 and the commands you gave was entered. But as you can see in the attached file, there is no difference

Vladamir, in the first attachment i posted, you can see that when i try to list the process that is owning the tape (show dev mga1 /full), the system said that there is no such a process. That is really silly..

As in the previous posting, i suspect the fibre_scsi patch and waiting a guy to verify my suspicion (so that i can call my local HP site and ask him why they did not apply fibre_scsi patch) or show me a point that i won't face with the same problem after reboot.

best regards...

[Vladimir Fabecic]

Forgot to say: please reboot NSR (or MDR) device which your tape unit is attached to. Then try sysman commands.

[mustafa_12]

if you mean Compaq msl5000 library device in which my tape drive is installed, i had already rebooted it after changes. but still there was no solution.

[mustafa_12]

By the way, i forgot to say that my node where i have problem is in a cluster environment (4-node). this may help to solve problem, i hope...

[Vladimir Fabecic]

MSL5000 can have NSR "card" in library unit, but also can be connected to external device. In first case you rebooted both NSR and tape, but in second case you have to reboot external NSR (network storage router) or MDR (modular data router, older device).
FC is not as simple as SCSI and causes more problems.

[Aaron Lewis_1]

mustafa, could there already be a device '2', perhaps in use by another node on the cluster?

[Volker Halle]

Mustafa,

the 'tape allocated to non-existant process' problem was supposed to be solved by VMS732_MOUNT96-V0100. This patch is included in VMS732_UPDATE-V0400, which you have installed.

So maybe there's still some scenario, which has not been fixed. If you can, try to force a crash and escalate this problem to HP.

Volker.

[Jan van den Ende]

Hi Mustafa,

if I read the problem correct, then you stopped with CTRL-C.

This interrups the current command, but does NOT stop the process.

If this process still exists, then THIS is "another user" still using the tape.

Try to re-connect to this process, and DISMOUNT the tape.

But if this process is already gone (as per Vladimir's advice), then follow the others.

Proost.

Have one on me.

jpe

[mustafa_12]

Aaron and Vladamir, I have changed the SCSI id to different numbers, I have rebooted both the library and the NSR (is it the one that connects a SAN switch and a library, one leg of which talks FC and the other talks SCSI?) and performed the sysman tasks again. But the result is fail again.

Volker, you are right. I have just informed the HP about the condition and waiting for an answer.

Jan, if the process that have access on tape existed, with what commands could I connect that process?

[Jan van den Ende]

Mustafa,

if that process still exists, and it is an interactive process (but how else could you CTRL-Y?) then log in with the username of that process.
The system will respond with: "You still have a disconnected process. Connect to it [Y/N] ?"
Enter Y and that is it.


Proost.

Have one on me.

jpe

[mustafa_12]

This is another environment that is similar as the problematic one. This environment was established to understant whay tape drive and OS behaves like that in the original environment. In the test environment, after I backuped up some files using tape drive, it goes like the one I described in my previous postings. But in this one there is a difference. The ACP process that controls the tape (MGAXACP) and the session process (that is shown as disconnected) goes RWAST state.

$ show user /full

SYSTEM XSERVER SYSTEM 21200441 TNA4: (disconnected)

but, if I logined xserver with telnet, I could not re-connect to disconnected session, or even OS did not ask a question to make an invitation.

[Willem Grooters]

Jan's approach to reconnect existing sessions depends on the configuration. Not all environments allow this, and will stop interactive processes if no connection to a terminal exist.

You can easily check this when you're behind a PC:

* Login to VMS
* Logoff from your PC without logout from VMS
* Logon onto your PC
* login to VMS again

In Jan's idea, you would now get the question to reconnect. In the enviroment I work in, the session is lost'.

Willem