SUMMARY: Scsi errors - mix of SCSI-2 & UW. (fwd)

From: Dan Kirkpatrick <dkirk_at_suhep.phy.syr.edu>
Date: Mon, 06 Apr 1998 16:43:08 -0400 (EDT)

Thank you:
aad_at_nw.verio.net
"Alan Rollow" <alan_at_nabeth.cxo.dec.com>
"William H. Magill" <magill_at_isc.upenn.edu>
System Prestidigitator <BOLSON_at_frango.hs.washington.edu>
Wesley Darlington <w.darlington_at_am.qub.ac.uk>
Mitch Bertone <mbertone_at_gtech.com> (thanks for ULTRA-SCSI msword doc)
Girish Phadke <PGIRISH_at_binariang.maxisnet.com.my>

SUMMARY (Original message at the end of summaries)
-------

Quick answer... I was exceeding length for ultra scsi.
I'm using single ended, not differential, which limits 1.5m for ultra wide.

>From the SCSI intro web page at
http://www.network.com/Products/Connectivity/TechBrief/SCSI/SCSIintro.html

SCSI Maximum Bus Length
                Single Ended Differential
                (Meters) (Meters)
SCSI-1 6 25
Fast SCSI 3 25
Fast Wide SCSI 3 25
Ultra SCSI 1.5 25
Wide Ultra SCSI - 25
Wide Ultra SCSI 1.5 or 3 25

You must also count at least a foot or more of cabling inside each
external enclosure.


One of the best, informative responses:
---------------------------------------
(from Wesley Darlington <w.darlington_at_am.qub.ac.uk>

SCSI comes in two flavours: differential and single-ended. These are
two totally separate branches of the scsi family. Differential scsi
allows long cable runs (>15m) whereas single ended scsi has a 1.5m
to 3m restriction. Note that UltraWide Differential and UltraWide
Single Ended are both possible and you can't share disks between the
two: a disk is differential or single ended, as is a scsi adapter.

Anyway, with ultra scsi, the chain can be 3m in length. In total.
This includes all internal cabling!!!

However, if you have four or more devices on the chain then the
length restriction suddenly halves to 1.5m. It's incredibly stupid.
It sounds like this is what is happening to you.

If your system is single ended then you have two choices: get 0.5m
external cables or get one big scsi box and put all your disks in it.
Note that even replacing *all* your 1m cables with 0.5m cables might
not solve the problem!

If you don't know whether your system is single ended or differential,
then chances are it's single ended. It's a bit like diesel or petrol
engined cars - if you don't know what yours is, then odds are it's
petrol! :-)

The latest scsi standard ultra2scsi uses a system called LVD (low
voltage differential) which allows long cable runs without the
excessive power consumption of traditional differential drives.
Also, AFAIK, one can plug a single ended scsi drive into an lvd
chain, but one can't plug in a traditional differential drive!

Regards,
Wesley Darlington,
Dept Applied Maths,
Queen's University of Belfast.

PS. Is is possible that the new 23gb drive is accidentally terminated?
This can produce some funny results too...



---------- ORIGINAL message ----------
Date: Fri, 3 Apr 1998 17:05:40 -0500 (EST)
From: Dan Kirkpatrick <dkirk_at_physics.syr.edu>
To: decstation-managers_at_ornl.gov, alpha-osf-managers_at_ornl.gov,
    support_at_adaptec.com, dean_at_adsys.com, kathyd_at_aspsys.com
Subject: Scsi errors - mix of SCSI-2 & UW.

Sorry if this is the wrong place to ask... didn't know where to turn...
Having a problem with a scsi drive/cabling... (will summarize responses)

Here's the setup that worked...

CPU--->23gbUW--->23gbUW---(68pin to 50pin cable)-->9gb fast scsi2--->TERM
         ID=2 ID=3 ID=1

We added another 23gbUW disk in line with the others... and quickly got a
lot of timeout errors from that new disk (ID=4).

CPU-->23gbUW-->23gbUW-->23gbUW--(68p-50p cbl)-->9gb fast scsi2--->TERM
       ID=2 ID=3 ID=4 ID=1


The errors we got were:
Apr 2 17:37:35 hepsu03 vmunix: cam_logger: CAM_ERROR packet
Apr 2 17:41:06 hepsu03 vmunix: cam_logger: bus 0 target 0 lun 0
(actually, repeated for target 0,1,2,3,4)
Apr 2 17:41:06 hepsu03 vmunix: ss_perform_timeout
Apr 2 17:41:06 hepsu03 vmunix: timeout on disconnected request

And on other machines trying to access it with automouting:
NFS3 RFS3_WRITE failed for server hepsu03: RPC: Timed out
NFS3 write error 60 on host hepsu03
NFS3 RFS3_READ failed for server hepsu03: RPC: Timed out

Now... each one of these cables are about 1 meter. Ultra should handle
15m.... Is the cable length combined with the narrow scsi causing the
problem? Why would it affect the last Ultra drive (ID=4)... I'd think it
would affect the narrow scsi.


Any ideas why? What if you have 10m of cabling for ultra wide, then stick
a narrow scsi-2 on the end that only has a length of 3m?
If you have a narrow scsi-2 on a chain full of ultra-wide, do you have to
limit the length to 3m?!

FYI for others... you CAN mix wide with narrow, provided you get an
adapter/cable that terminates the upper 8 bits, but I have no idea about
this cable length limitations.
A good cabling site is
http://www.adaptec.com/support/configuration/connect.html

Thanks!

- Dan Kirkpatrick - dkirk_at_physics.syr.edu -
| Systems Administrator |
| Physics Department |
| Syracuse University, Syracuse, NY |
-------------------------------------------
Received on Mon Apr 06 1998 - 22:44:57 NZST

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:37 NZDT