Me again! Tape drive device names this time

[Paul Condren]

Time for another problem!

On a couple of our servers, mainky 11.23 PA RISC when we reboot the server the /dev/rmt/0mn number change even though there have been NO changes to the device hardware paths.

These are fibre attached Ultrium 2 drived connected via an EMC Brocade fibre switch.
before the boot hey would be /dev/rmt/1mn, 2mn, 3mn and 4,n, then after a reboot they go to 5mn, 6md, 7mn and 8mn!

So we update the drive congig in data protector then the next time the server is booted they switch back again, or maybe they dont, but sooner or later they do.

Its causing no end of problems and Data protector leves tapes on the drives and fails to write data to them.

Can we stop this and force them to be stating by having a /dev/rmt/drive1, drive2 etc? I noce the minor number changes so I dont think this is possible.

they are IBM drives so we are using the latest atdd driver which riungs an /sbin/init.d script on server boot. could this be to blame?

ta

[Pete Randall]

There is no significance to the name. You can rename them to anything you want.


Pete

[Paul Condren]

Thats right, I can call them /dev/rmt/tom, dick and ahrry, but on a reboot they will come back as different minir number and will therefore no longer correspond to these names. Its almost like the HW path has changed by moving the fibre to a different card or switch port but I know for sure this is not the case.

[Torsten.]

I remember some similar threads here with the same scenario. AFAIR this belongs to a setting of the library. Sorry, I can't remember more details.

[Sp4admin]

Hi Paul,

We had the same problem. what we did was use the device files with the /dev/rmt/c5t3d0BESTnb and moved away from the /dev/rmt/31m and that seem to fix the problem. I'm not really sure why that fixed the problem however.

Hope that helps!
sp,

[Paul Condren]

Thats worth a try. Should the best drive names remain unchanged then? Ive not noticed.

I'll try a reboot and give it a go.

thanks.

[Rita C Workman]

I have to agree with Sp4admin. We have an EMC array and Brocade for our SAN - we have never had this problem by using the c-t-d- address for everything from library robotics to drives. And we just moved our entire data center and not even one address on any library (or server for that matter) changed.

Rgrds,
Rita

[Rita C Workman · Accepted solution]

Yes - use /dev/rmt/c-t-d-BEST or BESTn.

/rcw

[Duncan Edmonstone]

Paul,

Do you have 'before and after' examples of the hardware path?

Also can you confirm exactly how the tape drives/tape library are attached? (direct fibre tapes, or through a FC/SCSI bridge of some sort?) and what mode the tape library and drives are?

Sorry to worry you, but for another tale of woe with IBM SAN attached tape drives see here:

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=1042526

This never got fixed!

Incidentally we never seem to have these kind of problems with HP tape drives!

HTH

Duncan

[Paul Condren]

Hi.

The server models bvary but the one I'm trying to resolve this on (because its easier to tge a reboot slot!) is an RP4440 running 11.23 on PA-RISC. It has a fibre connection to a 32 port EMC brocade fibre switch where it can see 4 fibre IBM ultrium 2 drives which are housed in a StorageTek L700.

The harware and infrastructure has not been changed for a while and we never used to have this problem. However with the introduction of 11.23 and hence a newer version of the IBM atdd driver this started to happen. This is a dev environment but we are also having this problem in our live/production environments which siimilart setup.

The server I am testing on is down today with a boot disk failure. Once its back I'll test some more with some reboots and try and get a before & after view for you.

thanks.

[Torsten.]

On the L700 I know there was a fibre to scsi bridge in use. Maybe you can check and adjust something in there.

How about the hardware path to the drives?
Is something different here?
Maybe it is mapped to another LUN or something. This may cause the OS to "think" it is a new device and it will create new device files.

[Paul Condren]

OK, my test server is back again now. Below you can see the results of my initial testing today, all I did was ioscan for tapes, then shutdown -ry 0 and the server comes back up and the path has changed! this happened on every reboot. Meanwhile the L700 is static. We dont have and SCSI ro fibre bridges - the drives are fibre straight to the switch.

The good news is it looks like the 'BEST' device name remains unchanged.

File attached...

[Torsten.]

IMHO this must be a driver issue.

I never seen this with the original HP-UX tape driver (stape), but I guess you are using another.

Ask the driver vendor!

[Paul Condren]

Torsten - correct - these are not HP drives but IBM ones so they dont use stape.

Ive now done an additional test which seems to have stopped the flipping - I have removed the link in /sbin/rc2.d to the atdd script for server boot. After this the drives do not change any more.

Looking at this script it seems to be doing the IBM/atdd equivelant of an insf -e etc to rediscover the drives on each server boot. THis is not necessary unless you add new drives in which case you can run the command manualy. I dont see any need for it to be run on each boot.