[comp.sys.ncr] braindead

TEMNGT23@ysub.ysu.edu (Lou Anschuetz) (11/26/90)

This is sort of a followup to my prior question about my often
braindead tower 32/700.  It did indeed die the same way three times
over the holiday weekend, requiring a 140 mile drive at one point to
come back and fix it.  Again, no error message, just a login prompt
and no response.  This phenomenon appears to be related to some
problem with getty's, but is impossible to diagnose.  Any help
would be appreciated.  Second, we are considering buying a terminal
server (16 port) to do away with HPSIO systems.  Anyone have a good
(cheap) recommendation on one that will perform dedicated access to
to the tower (ie: a dial in to the server should automagically
connect you to the tower and not give you an annex prompt only).

Second, prf is a mystery.  I can create a file called /dev/prf,
which does not exist and thus allows no use of the profiler.  With
this done prfld complains that no value is set for prfmax (and
indeed there is no parameter for this in the makefiles), and
prfstat on complains that there is a -1 return code from the kernel.
BTW, prfstat on also reports that it turns ON the statistics.

Third, I have the normal 380 NBLK16 assigned.  From time to time
I get a report of 10 or 11 failures on this value, even though
the max used is never over about 285.  This may be a dumb question,
but how can this be?  This is even more puzzling since there are
plenty of spare 64byte blocks which never fail?

Finally, netstat has become completely broken.  If I am real lucky
it reports 1 second of time used, but never returns (after printing
the column headers).  Most times it reports 0 time used.  Is there
something besides netstat (I have reloaded it from the install tape)
and /tmp/netstatdata that is required for this program to work?
I know that this was a problem in 4.00 of win tcpip, but I thought
it was fixed in 4.01.  Indeed, it was working correctly up until
last Tuesday, though it was still terribly slow for a function that
should be reading a memory value....

Sorry if some of this sounds angry, I'm just tired from driving in
every night over the weekend to power off the machine, power it on,
say no to dumping to tape, powering off, powering on, going into
single user, fsck, and finally getting up and going.....  :-(

Lou Anschuetz
temngt23@ysub.ysu.edu
root@yfn.ysu.edu

TEMNGT23@ysub.ysu.edu (Lou Anschuetz) (11/26/90)

A little further information is now available.  If I physically
turn off the modem, it will cause the line to reset.  My guess is
that this turns off dsr, but a simple line hangup is not read by
the tower as a dropped line.  Of course, this is difficult to do
from 140 miles away, and inconvenient if closer.  Has anyone solved
this problem with processes that won't die on a getty line with
any simpler techniques?

nick@aimed.UUCP (Nick Pemberton) (11/29/90)

In article <90330.082142TEMNGT23@ysub.ysu.edu> TEMNGT23@ysub.ysu.edu (Lou Anschuetz) writes:
>would be appreciated.  Second, we are considering buying a terminal
>server (16 port) to do away with HPSIO systems.  Anyone have a good
>(cheap) recommendation on one that will perform dedicated access to
>to the tower (ie: a dial in to the server should automagically
>connect you to the tower and not give you an annex prompt only).
>

I highly recommend the DTC/RTC system that NCR supports (I believe it
is from SYSTEK) - this is an arcnet based system. We have three machines
with this system and it is far superiour to the HPSIO system. You can
connect 128 serial ports and 8 (?) parallel ports per DTC controller. On
our main machine we have 80 ports hanging on one controller, and it has
never had any trouble.

Nick.
-- 
Nick Pemberton  		 uucp: !{lsuc, uunet!mnetor}!aimed!nick
AIM, Inc			  bus: (416) 429-1085
Toronto, Ontario, Canada         Home: (416) 690-0647

vernon@cs.utexas.edu (Vernon E. Hurdle) (12/01/90)

> 
> A little further information is now available.  If I physically
> turn off the modem, it will cause the line to reset.  My guess is
> that this turns off dsr, but a simple line hangup is not read by
> the tower as a dropped line.  Of course, this is difficult to do
> from 140 miles away, and inconvenient if closer.  Has anyone solved
> this problem with processes that won't die on a getty line with
> any simpler techniques?
> 
This sounds like you have a cabling or modem  problem.   Here  is
the  pinouts for a modem cable that I use for all different kinds
of modems and models of Towers.


                DB15             DB25
		  1 -------------  2
                  2 -------------  4
              |-- 3 -------------  6
              |   4 ------------- 20
              |   .                .
              |   9 -------------  3
              |  10 -------------  5
              |  11 -------------  7
              |--12 -------------  8

If you need a DB9 instead of a DB15 just replace the numbers 9-12 with
6-9 respectively.

Secondly, what kind of modem do you have?  Many modems have reset
options  built  in  that  can send out the equivalent of ATZ upon
termination of the connection.  Check the modem's manual.

I hope this is of some help.  --

#################################################
# Vernon E. Hurdle (vernon@ssi600.lonestar.org) #
# Seay Systems, Inc.   Dallas, TX               #
# Voice: 214/522-2324                           #
#################################################