disk timeout, SCSI reset...

[Bernard J. Duffy]

These kinds of "disk timeout" SYSLOG messages happened on our 4D/25
4D/220 systems and the only thing that stopped them was swapping out
some cables and in one case changing the SCSI bus device ordering.  Do
make sure you have no (internal, since it's real hard to have
external ones in the middle of the bus ;-) ) terminators on the devices
in the middle or the SGI disk drive inside the power/disk tower/cabinet.
After you make sure of that, use the shorter and shorter cables.  It
might help to do a quick test with the shortest cables you have and work
up from there as space and cabinets require (can't use a real short
cable out of the PI boxes with the case on it).  You can also try
some tests with different arrangements of the SCSI bus.  In one case
I had to put the 8mm Exabyte on the end of the bus in order to get
things working quietly.  
    Not all of these disk-timeout errors aren't fatal, they just put 
a nasty pause in I/O.  For each of your arrangements/cableing, fire
a tar/bru read/write operations.  This should run error free/ pause
free on a lightly loaded system.  I usually fire off two tar/bru
jobs to separate drives on the same SCSI bus for my final verification
test.  This is an extreme test since most daily operations are not
this SCSI I/O intensive.  Our 4D/220 will occasionally pause under
this test, but the I/O operation continues and bru operation doesn't
fail (I used bru's check/verification option).

Good luck,

-- 
Bernie Duffy   Systems Programmer II | Bitnet    :  BERNIE at UMBC2
Academic Computing Services - L005e  | Internet  :  BERNIE at UMBC2.UMBC.EDU
Univ. of Maryland Baltimore County   | UUCP      :  ...!uunet!umbc3!bernie
Baltimore, MD  21228   (U.S.A.)      | W: (301) 455-3231  H: (301) 744-2954