Errors with mc command

[Alan Riggs]

I'm not sure where to put this wuestion, so I'm dropping it in familiar waters. The situation:

ADIC Tape library: a tape had jammed during retrieval from the mailbox and then fallen to the floor. I popped open the front and reslotted the tape. No problem, right?

Well, in teh past I have had issues with the door not being "really shut" so that while the initial barcode scan would function properly nothing else would. I use mc commands to test library operations when that happens. No problem, right?

I wish. While these commands have functioned for me in the past, I now receive:

# mc -p /dev/picker -r DS
ERROR: 0x5 Illegal Request : 0x2400 Invalid field in command descriptor block

Confused by this, I tried using uma to do some tape manipulations:

# /opt/omni/lbin/uma -ioctl /dev/picker
*** PROGRAM: UMA VERSION: HP OpenView OmniBack II A.03.50

*** Copyright (C) 1996 Hewlett-Packard Company
*** License is restricted for use with licensed
*** HP OpenView OmniBack II products.

/dev/picker> stat
'Element Status (T=Transport, X=Im/Export, D=Drive, S=Storage):
1 T1 Empty "" ""

16 X1 Empty "" ""
.
.
.
/dev/picker> move from 4182 to 16
move: Invalid element address.

dev/picker> move from S87 to X1
move: Invalid element address.

So, door not seated right? Nope. Library self-checks and direct manipulations work fine. And, here's the kicker, I can load and initialize tapes successfully through omniback! So, am I missing something in the mc/uma commands? I admit that I don't use these evrfyday, but I can't see the errors (invalid field/invalid element) the error message is referencing. Any ideas?

Now, it has been a very long start of the week

[A. Clay Stephenson]

Hi Alan:

This is rather puzzling. One thing you might try is the uma 'doorlock 0' to unlock the mailslot door and 'doorlock 1' to lock it.
It might require these commands before any 'move' commands are issued.


I assume that your uma 'stat' command indeed shows that S87 is valid. I suppose that it's also possible that you have some OB2 processes hung (possibly on other hosts) - or hung mc commands.

[Alan Riggs]

It is puzzling, but it appears to be patch related. PHKL_22460 (cumulative SCSI) apparently broke some ioctls that send null data (like mc). It is corrected in PHKL_22759 and later. I will try the patch and see whether it resolves my problem.

If not, then I will be well and truly stumped.

[Angus Crome]

I know it is rather plebian, but have you tried the old standby of power/cycle the library and reboot the cell-server?

Another possibility is that your picker is going out. It happened to me on my HP branded StorageTek after about 4 years of service. Got the exact same errors with uma and mc. Eventually, they started showing up in omniback as well.

I think your idea of the patch is definitely the proper place to start though.