Jeremy,
I have also seen this same error. My environment:
ADSM V2.1.5.11 running on a RS/6000 7013 58H with AIX 4.1.4
The RS/6000 is attached to a 3490E B40 tape drive via 2 parallel channels
using S/370 channel emulator/a adapter cards.
Device drivers are provided by PRPQ '5799-QDA Parallel Tape Attachment/6000'
I configured the 4 drives in the B40 as follows:
unit address 500 = /dev/rmt1
unit address 501 = /dev/rmt2
unit address 508 = /dev/rmt3
unit address 509 = /dev/rmt4
I made several changes to my configuration with the following results:
1.
I had both channels defined to the A20 control unit in front of the B40 with the
first 2 drives defined to ADSM: /dev/rmt1 ,2. I always seamed to get the error
on the first drive (/dev/rmt1). When I got the error, it would not recover, it
would hang the drive.
ADSM would ask for a mount on its activity log but no mount was displayed on the
tape drive. You could mount the tape on the drive where ADSM wanted it but it
would never see it. It kept asking for the tape until the mount request timed
out. At this point, if I tried to use /dev/rmt2 I would also get I/O errors. To
free the drives I had to stop/start ADSM and in some cases delete, re-add and
configure the drive to AIX.
We also put on the latest microcode with no improvement.
2.
I was not sure that my configuration was valid so I removed one of the channels.
Now when I used the same 2 drives (/dev/rmt1 , 2) I saw the exact error that you
described. Sometimes during a single migration from the disk pool to tape pool I
would get many I/O errors and therfor many tapes set to R/O. (But atleast it
would recover from the error)
3.
I deleted drives /dev/rmt1,2 from ADSM and defined /dev/rmt3,4. I have not seen
the error since. I am still only using 1 channel. I am now trying to find out if
2 channels to the same A20 is a valid configuration from the RS/6000. Internally
the A20 has 2 logical control units and I had 1 channel to each. It's possible
than an I/O can go down 1 channel and the interupt returned on the other.
Normally the operating system can handle this. I'm wondering if AIX sometimes
misses an interupt, so it keeps waiting and waiting for a response which results
in a hung tape drive.
Larry Robertson (810) 737-7300
larry_robertson AT compuware DOT com
______________________________ Reply Separator _________________________________
Subject: Re: (U) ADSM I/O error ANR8311E
Author: "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU> at
CWUS-Internet
Date: 4/23/97 2:41 PM
Jeremy,
I get this very infrequently on our 3490Es and although the ADSM message &
the AIX errlog point to the drive as the problem, I know it's not true
since ADSM continues to use the drive in question without any other
problems. Therefore, I have simply treated this as a bad tape situation and
moved the data out of the volume. Also, I'm pretty certain it's not a
end-of-tape situation because the response from a 'q vol ...' shows the
volume as filling & less than 100% util.
At 12:50 PM 4/23/97 EDT, you wrote:
>We are running ADSM v2r1 on AIX 4.1.3 with IBM 3490E tape drives.
>
>Often, we get the above error during a nightly backup, upon which ADSM
>sets the affected volume to readonly (ANR1411E) and moves to the next vol
>in
>the storagepool (ANR1411W). Although the error is reflected as a TAPE_ERR
>in the AIX errlog, we are pretty sure it is NOT a genuine hardware
>problem - we have changed every hardware component and microcode level.
>
>Our current suspicion is that another signal (possibly EndOfTape?) is
>being misinterpreted and producing this error.
>
>Does anyone have any experience of this sort of problem?
>
>Thanks,
>
>---------------------------------------------------------------
>Jeremy Worrell, IBM UK Ltd. x.664884
>Beauchamp Sophisticate ext. 01926 464884
>---------------------------------------------------------------
>
>
Have a nice day or whatever's left of it.
David Ong
National Semiconductor Corp.
|
