[Info-vax] DEC 3000/900 bootup problem

[Vance Haemmerle]

John Santos wrote:

> In article <Qy07l.16522$ZP4.1997 at nlpi067.nbdc.sbc.com>, 
> vance at toyvax.Glendale.CA.US says...> 
> 
>>SMS wrote:
>>
>>
>>>Vance Haemmerle wrote:
>>>
>>>
>>>>Today my DEC 3000/900 developed a problem.
>>>>It was in a cluster with a VAXstation and was
>>>>booted out of the cluster, both had been up for
>>>>42 days, it appeared to have hung.  Now the DEC 3000/900
>>>>cannot boot.  [...]
>>>
>>>
>>>   Sounds bad.
>>>
>>>
>>>
>>>>Whether booting from the system disk, an internal
>>>>RRD45 or a newly connected RRD43 (just in case it was
>>>>a disk or cd-rom problem) after all the inits and
>>>>book checks pass and the
>>>>
>>>>OpenVMS (TM) Alpha Operating System, Version 7.1-2
>>>>
>>>>message appears and a few more disk activities occur
>>>>(drive or CD-ROM) the bootup appears to hang.
>>>>
>>>>
>>>>Details and things I've tried:
>>>>
>>>>All power-on tests are OK
>>>>All console tests pass
>>>>The number on the console display is normal ("90").
>>>>A boot from the V7.1-2 and V7.2-2 CD-ROMs hang
>>>>soon after the OS version string.
>>>>Conversational boots with "MIN", USE DEFAULT and SET/STARTUP OPA0:
>>>>all hang.
>>>>The system also has a drive with Tru64 Unix V4.0F on it
>>>>but it boots all the way up and runs!
>>>>
>>>>Any ideas of what to try or what it might be and why it only
>>>>affects VMS bootup?  I assume something in the autoconfigure
>>>>is going bad.
>>>
>>>
>>>   Is it using the same console as you are?  (And what is
>>>that?)  In some cases the early messages get sent to
>>>multiple ports, but then it narrows its scope to one, and
>>>if you're using the wrong one, it looks dead.
>>>
>>>   If it's trying to cluster, does a REPLY /ENABLE
>>>terminal on another system show any OPCOM messages?
>>
>>   By console, I just mean the attached keyboard, and local
>>">>>" display (I'm using it as a Alphastation).  OPA0: on
>>the VAXstation is the cluster console.  No messages appear there
>>during boot attempts of the Alpha, I do NOT get to the message
>>on the Alpha: "waiting to form or join a VMScluster system".
>>The boot hangs before that message ever appears.
>>
>>   Booting from a CD-ROM should not have it try to cluster.
>>I've also done a boot -fl 0,1 and when the SYSBOOT> prompt
>>comes up I've SET VAXCLUSTER 0.  That does not boot either.
> 
> 
> I think Steve's point is the console variable may be set
> incorrectly, but you might not notice this until too late...
> 
> It sounds like you're using the graphics console (VGA screen
> plus keyboard and mouse plugged directly into the Alpha.  If
> the console variable has gotten set to serial (VT or LA terminal
> plugged into an RS232 DB25 port, an RS232 DB9 port or an MMJ
> port on the Alpha depending on the model, I forget what a
> 3000-900 has), it might look normal until VMS decides it needs
> to ask a question and is prompting for input on the wrong console.
> 


   This is true, I'm using the graphics console (in VGA mode)
plus the keyboard plugged in directly to the Alpha.


> At the >>>, do
> 
>   >>> show console
> 
> and
> 
>   >>> set console graphics
> 
> if it shows as serial (not sure if it can be anything else.)


   The DEC 3000/900 is older than this console environment
variable.  I also have an XP1000 which you can set console but
the 3000/900 just says "?23 ILL CMD" when you type "SHOW CONSOLE".
This did give me an idea to try another thing.  I set the S3
switch in the back to the alternative console setting and set my
printer to 9600 baud (it is hooked up to the console/printer port).
I was able to get a printout of the attempted boot (after setting
the reset button to get enough ">>>" to fill a page.  But no
additional output is seen that what is seen on the VGA display.


> 
> If the CMOS battery has died, it might have lost some or all
> the console configuration, causing this.  It may have also
> lost the time, so it may be specifically be asking you to
> enter the date and time on the serial port.  If so, all you
> need to do is wait 1,000,000 microfortnights and it will
> continue with its best guess about the current time :-)


   I think this may be on the right track.  I have had this system
for over 12 years and never replaced the CMOS battery.  The
bootups from CD-ROM (and after I boot Tru64 which has an
incompatible usage of the TOY clock than VMS) would always
request the current date/time when booting up.  I do not see
these messages to set the time.  I have tried entering the
time (e.g. "01-JAN-2009 14:02") after the apparent hang but
that doesn't do anything either.

   After my attempt with the S3 setting, the console did ask
me to set the language again, but other console settings, e.g.
bootdef_dev, fast_scsi_b, scsi_reset, etc. still keep their
non-default settings.

   Would the NVR (manual says "Non-volitile RAM and time-of-year
(TOY) clock") test pass even with a dead battery?  The
NVR test does some things, "ASSURE_CLOCK_IS_TICKING TEST"
and "INTERRUPT TEST" all of which pass.  Would a dead battery
cause a system to stop?

   The system does appear to hang at the point where it would
ask the system time.  Anyone know of a hardware problem
which would cause this but also have the system pass all
hardware tests?


> 
> Or the console setting might always have been wrong, but
> you never noticed before because it never before needed
> the answer to a question during the boot.
> 


   The configuration and console settings have not changed
for years.  In the past years I have booted UNIX (not for
a while though) or from a CD-ROM and have set the time
before.  Like I said earlier, the serial console is set
by the S3 switch for these systems so I don't believe it's
possible to lose a console message due to a wrong setting.

--
Vance Haemmerle