Adding a 3rd party SCSI disk to an rs6000

[Alan Stebbens]

In <1149 at cameron.egr.duke.edu> jpe at egr.duke.edu (John P. Eisenmenger) writes:

>I'm trying to figure out how to set up a WREN-VII for my RS6000.  Has
>anyone done this before?  The problem, as I see it, is that I need to
>define a new disk subclass in order to get little things like the disk's
>size correct.  How do I go about it?

>I'm also having some troubles with a SCSI 8mm tape drive.  It looks
>like its configured (marked as available, correct subclass, and device
>files are there), but whenever I try to access it I get:

>  /dev/rmt1: A system call received a parameter that is not valid.

About the SCSI disk drive and its configuration:

First, I don't believe that you need to add a new disk subclass.  Using
SMIT, I just added a new disk, using the "Other SCSI" as the disk type,
then I added a new physical volume as well as a new volume root, since I
wanted to be able to export this entire volume set.  Using "Other SCSI",
the system seemed to know how to read the SCSI parameter table from the
disk drive and determine its size. 

Once you create a physical volume, you can either (a) add a new filesystem
(which will automatically create logical volumes for you), or you can (b)
create logical volumes and then create filesystems using "previously
configured logical volume".  I did (a) first only to discover that the
default logical volume had a default name which I disliked, and more
importantly a default size which didn't match the available size of the
disk.  It seems that the logical volume naming convention differs for
the pre-installed systems from that used by default for SMIT.  I like to
have all logical volumes named consistently, so I ended up having to do
it by hand. 

One thing which is missing from SMIT (at least on AIX 3.1), is the
capability of changing the JFSLOG volume for a filesystem.  I ended up
having to discover how to do this by hand, by searching the scripts in
the /etc and /usr/lpp/inst directories.   This was so I could have the
logical volume name used for the journal logging on my new hard disk be
consistent with the name used for the other disks.

As for the device errors:

I have an RS6000/320 with three external SCSI devices attached on the
same "scsi0" controller with two internally attached 320MB drives.  If
you read the install instructions for the SCSI controller, you'll note
that you're not supposed to be doing what I've done.  If you want more
than two external devices AND two internal drives are already connected,
you are supposed to purchase a second SCSI controller with which you can
attach up to seven devices.  Now that that has been said... 

I had this same problem on my new unit, which led me to RTFM and
discover the above "limitation".  I was above to give up, when it
occurred to me that the accessibility of each device may depend upon the
order in the chain, if the SCSI driver were at "its limit" in driving
the SCSI bus signals. 

The SCSI cable from the CPU has piggy-back connector at the device end, which
I'll label cable "1", with ends "A and "B".  The "B" end has the piggy-back
connector receptacle.  The "A" end is the usual Centronics style "pinch-slot"
connector.  IBM also supplies a shorter SCSI daisy-chain cable with similar
connectors: an "A" end, and a "B" piggy-back connector end.

My first attempt at connecting all the SCSI drives was as follows:

	CPU [A]-----(1)------[B] 1/4" tape drive
			 ----[A]
			(2)
			 ----[B] CD-ROM
			 ----[A]
			(3)
			 ----[C] 720 MB SCSI Drive (Hitachi)

Notes: Cable (3) is a "home-built" ribbon-cable with a Centronics style 
(pinch-slot) at end [A], and a 3x20 (DB60?) pin connector at end [C].
The Hitachi drive has an internal SCSI terminator, so no external terminator
is needed.

Anyway, the configuration above partially worked: the CPU could access the
1/4" tape drive, and the disk drive, but not the CD-ROM.  So, I tried the
following configuration, which is what I'm successfully using now:

	CPU [A]-----(1)------[B] CD-ROM
			 ----[A]
			(2)
			 ----[B] 1/4" tape drive
			 ----[A]
			(3)
			 ----[C] 720 MB SCSI Drive (Hitachi)

In the unsuccessful configuration, the error messages and symptoms were 
exactly the same as you describe above.  I can't say if these are the only
reasons for such errors, but you just may want to try changing the order
of your SCSI configuration.  

Remember, also, that both configurations are NOT SUPPORTED by IBM, if you have
internal SCSI drives on the same SCSI controller.

And, as always, your results may vary ... :^)

Alan Stebbens   Computer Resource Manager
		Center for Computational Sciences and Engineering (CCSE)
		University of California, Santa Barbara
		3111 Engineering I
		Santa Barbara, CA 93106

Internet: aks at hub.ucsb.edu
BITNET:   aks%hub at ucsbuxa.bitnet
UUCP:     ...{ucbvax,sdcsvax,cepu}!ucsbcsl!aks
Voice:    (805) 893-8135 (CCSE Office: 893-3221)

--

Alan Stebbens <aks at hub.ucsb.edu>