Subject: Re: Weirdities (was: Re: a question about bwtwo1 (cgfour))
To: Jari Kokko <jkokko@cc.hut.fi>
From: John Little <gaijin@doemail.sbi.com>
List: port-sun3
Date: 05/28/1996 07:56:26
DISCLAIMER: I'm assuming that most of the people on this list have a 
	    reasonable familiarity with Sun hardware. If you don't, you
	    should not attempt the type of maintenance I describe
	    here. See the warning below.


Jari,

   Multitudinous wierdness like this is a dead ringer for a bad
   power-supply connector. Take the system (plastic) top cover off
   (undo the two small screws at the CPU side and then lift from that
   side and unhook from the back). Take the metal top cover off the PSU
   (sometimes this has a screw, but on most models it is an
   interference fit; it just slides out). Check the push-fit connector
   where the PSU plugs onto the backplane.  Discolouration of the
   plastic indicates heat, which means that voltage is being dropped
   where it shouldn't be.

   Clean the pins on the backplane side with a wire brush until they
   lose that dull, "cooked" look. You might be able to clean and
   retension the connectors inside the plastic housing (PSU side) if
   they haven't cooked to the point of being brittle and if you have
   the tools necessary to extract the pins from the connector housing.
   The best option is to replace the whole connector assembly, or the
   complete PSU unit.

   *WARNING* - The procedure described above brings you into contact
   with components normally operating at mains voltage.  You *MUST*
   disconnect the system from the mains power before attempting any of
   the maintenance described above. If you are not familiar with
   switching power-supplies and equipment operating at mains voltage,
   DO NOT attempt to carry out these procedures yourself, but take the
   system to a repair shop. If you kill yourself trying to fix your
   PSU, I'll never try to help you again, okay?!? :-)

   Oh, and the PROM diags? Do trust them. The PROM monitor is your
   friend, believe me!


		Good luck!


			-John-

-------
% 
% Well, since I wrote that, I tried booting diagnostic, and then some
% weird things started to happen: 
%   -I got one LED error pattern, which I think was interrupt controller
%    something error
%   -I got a normal boot, and got a running system for a while, then
%    it hung when I was pinging a host (ppp), but the LEDs kept walking
%   -I got an MMU fault panic
%   -I got parity errors from si0, fscking, system hung
%   -I got a non-diagnostic boot to work
%   -I got a non-diagnostic boot, which left LED 1 on, the others,
%    except LED 0 dimly blinking
%   -I got a memory error LED pattern booting DIAG
%   -I got a crashed system, which panicked reporting error in CPU cycle
%    and said that all bets were off (which I knew :-)
%   -plus various other things which I can't remember
% 
% With such clear diagnostics, it's easy to say that _something_ is
% definitely wrong :-)
%