Re: TASS UPS, Tech Note.

[Tom Droege <droege@FNAL.GOV>]

Chris writes below.

Yep Chris, there are two groups.  Actually there is a third group
that builds systems even more primative than DOS.  Though we seem
to be a dying breed.  There seems to be an OS in my Waffle maker
these days.

I think each has it's place.  I would do the real time running with
a simple machine, say DOS based as we are doing.  The data processing
is much more complex, and we want to be able to restart a data base
where it was, with all the updates carefully protected to make sure
we know exactly what state it is in.  On the other hand, if the real
time machine crashes, we have lost that data block anyway.  No sense
worrying about it, or trying to restart where we left off, as the 
sky is gone.  I would not try to do both on the same machine.  It is
asking for trouble on the real time process no matter how fast the 
big machine is.  Besides, old PCs suitable to run a camera are cheap.
I have several in my house that are too old to do anything useful 
except possibly run a prom burner. 

Looks like I will try to communicate between the two processes with 
something simple, like an Ethernet connection, where the big 
carefully protected machine is in command.

Tom Droege

At 12:03 PM 8/21/97 -0700, you wrote:
>The reset thread is kind of interesting.
>
>There seem to be two groups of people.  Those who learned about
>computers by using low-end PCs, the 8080, Z80 then later the PC
>and now Pentiums.  And the second group who learned about computers
>like the IBM 360 or CDC Cyber6400 then first saw UNIX on PDPs and VAX
>in the late 70's then started messing with PCs and other micros.
>
>
>To this second group of people the way one shuts down a computer
>is to first post notices on company bulliten boards saying "The
>XXX system will be shutdown next Tuesday from 1:00am to 3:00am
>to install a new set of disk drive.  Sorry for any trouble this
>may cause."
>
>Comming from the other side, I can see why people with experience
>with PCs and other micros would think a re-set would solve most 
>problems.  Because on a DOS machine it does.  99% of the system
>crashes on a DOS/Win machine are caused by buggy aplication software
>or a bug in the OS.  A reset will and does fix this kind of problem.
>
>On other systems applications simply don't (can't) crash the OS
>and the OSes themselve are fairly stable.  On these larger systems
>crashes still happen but are more likely caused by hardware failures.
>In this case hitting a reset button is not so usful.
>
>My experience with Linux is that the OS itself does not crash but
>the machine can get itself in a state where it stops  talking to
>the console keyboard.  Fixing this cleanly requires access to a
>second console or a remote login from a networked computer.  If
>the (partialy hung) Linux pc has only one physical terminal and
>is not networked then the only option is the reset switch.  Most
>UNIX machines are networked so it is pretty common for system admin
>types to do remote repair work.
>
>John D. Gwinner wrote:
>> 
>> Herbert:
>> 
>> > A more "robust" reset would be to shutdown the PC's power supply
>> > via an AC disconnect. It would have the additional virtue of shutting
>> > down the AC supply - sometimes they "hang". I had suggested a
>> > mundane AC daily timer control that would shutdown the PC *every day*,
>> > an assured restart at a convenient time resolves a number of problems.
>> > Some may not like this solution, but it costs $10 and is fairly idiot
>> > proof, and it saves energy if you power down for a signifigant period.
>> 
>> I really don't like this suggestions, as the 10 dollar timer has no idea if
>> you've got files open at that time.  You are practically guaranteed to
>> loose open files at some point.  A watchdog timer, or periodic (software)
>> reset has the same problem of course, but with the advantage that the only
>> time the watchdog will reset is if the machine has hung, not when it's
>> humming along fine.  A software reset will gracefully close file handles,
>> tell applications there time is over, etc.
>> 
>> In the days of multi-tasking operating systems, I think it's very very
>> dangerous to just unilaterally power off a machine.  Powering off a UPS if
>> the UPS connects to the serial port to tell the PC to shut off would
>> probably work OK (assuming that UPS monitoring software is available and in
>> use, which isn't tough).
>> 
>>                 == John ==
>
>-- 
>
>--Chris Albertson
>
>  chris@topdog.logicon.com                Voice:  818-351-0089  X127
>  Logicon RDA, Pasadena California          Fax:  818-351-0699
>
>