[Beowulf] Re: Emergency Power Off
mathog at caltech.edu
Fri Mar 23 12:17:21 PDT 2007
Jim Lux wrote:
>(with the icky aspect of some UPSes
> requiring active voltage to shut them down.. WHAT were they thinking?)
No kidding. Not only that, but when dealing with these UPS interfaces
a degree in mind reading and reverse engineering is required, since
the companies never publish full specs for the wiring on these ports.
The APC unit in the room provides its own 24V (more
or less) on one pin of the serial interface. That's convenient. What
isn't convenient is if you use their little EPO adapter the UPS can't
run "SMART" anymore (according to their tech support). So you lose the
ability to carry on most types of data transfer with the unit. Instead
it goes to "simple" (contact closure) mode. The APC has two different
shutdown pins, one causes it to send a shutdown command to the attached
PC, hang around a while, and then turn itself off. The other is an EPO
appropriate "turn off instantly". Now the "SMART" cable doesn't
actually use the EPO pin, or the DC pin, so one could wire in an
adapter between the SMART cable and the UPS that would
conditionally pull the poweroff line up without affecting the
normal "SMART" operation. Except, APC tells me this is an
"unsupported" mode of operation. One has to wonder why they didn't do
it that way themselves, so that their adapter could retain "SMART"
The TrippLite that takes a Y adapter cable is even more interesting.
The specs for that cable say that one need only short the center two
RJ11 pins to kill the inverter. Yet from everything we know about that
interface, the inverter kill line must be pulled high in order to kill
the inverter. So yesterday I measured the voltages on all the serial
port pins, and guess what, there's no significant positive voltage in
there. There is however -6.5V on one pin, so maybe the kill works if
that pin is pulled below ground as well? The only way to find out what
that cable is really doing is to buy one and analyze it.
Personally I find the way the UPS companies hide information about
the interface jacks on their UPS units to be super annoying. There
isn't enough engineering info available to make reasonable design
decisions about what you're going to hook up to them, unless it's just
the one computer to the one UPS, in which case the cable that came with
the unit is fine. Simple nontechnical instructions like "short
these pins" don't tell me enough to know what all the valid
options are. For instance, could an opto-isolator go in there?
What will the voltage be across those pins before it is shorted?
What is a "short" in this context: 0.1 Ohm, 10 Ohm, 1000 Ohm?
How much current could flow when the pins are shorted?
Little bitty relays can sometimes be welded shut by pretty small
currents. How much current must flow for the unit to turn off? As
things stand now the only safe thing to do is to put a pretty big
physical contact switch (perhaps relay driven) on it and hope they've
designed things so that the RJ11 pins don't melt in the jack when the
switch is thrown. Either that or I have to spend time experimenting
with the interface, plugging in resistors, measuring voltages or
currents - none of which should be necessary.
mathog at caltech.edu
Manager, Sequence Analysis Facility, Biology Division, Caltech
More information about the Beowulf