[Nut-upsuser] Sweex 1000VA UPS
Arjen de Korte
nut+users at de-korte.org
Thu Nov 9 21:04:09 CET 2006
>> Are you absolutely sure that the UPS is connected to /dev/ttyS0? For
>> 'upstype=7' the signals for 'on battery' and 'low battery' are -CTS and
>> -DCD (both zero). I wouldn't be surprized at all if this is the same as
>> nothing connected to the port you're monitoring (although I can't test
>> that right now).
> The machine kept rebooting, even with no serial cable to the UPS
> connected anymore, so I guess that's a yes?
I would say that if it doesn't make a difference whether the serial
cable is attached or not, that would be a safe bet, yes. I just checked
with genericups monitoring a serial port without anything connected and
as expected, it showed LB OB (low battery and on battery), which would
force a system shutdown had upsmon been running:
# upsc myups at localhost
ups.status: LB OB
(Shame on me, even I'm not using the lastest NUT version)
> How can I be absolutely sure it's connected to /dev/ttyS0?
I don't know. The trouble with contact closure type UPSes is, there is
no way to determine if the correct UPS is connected and in many cases,
no way to check if something is connected at all. Short of a couple of
ones that support serial PnP (I know of at least APC has a model that
does), but that's not used by NUT (should be handled at a different level).
> That proliant server has a serial A and a serial B RS232, I took serial A.
If you have a serial loopback plug (or a breakout box), you could check
if you can access that port, but looking at your past replies I fear
that you even don't know what I'm talking about right now (no pun
intended). I'm not familiar with proliant servers, so I can't offer much
more help here.
>> Lesson #3: You really, REALLY, shouldn't experiment with a UPS on a live
>> system... :-)
> Yes, Easy to say when the system is already operational. I didn't have
> that much choice.
Disconnecting a UPS from a live server and testing it at a separate
workstation is a MUST if you're not absolutely sure the configuration is
correct. I would never use an UPS on a production system without double
checking that monitoring works. Otherwise your investment gives you a
false sense of security.
Bottom line is that a UPS should make a power loss to your system
predictable, it can't prevent it always, so you MUST check if that
works. Even when monitoring the status of the UPS works, you must still
make sure the system shuts down cleanly and reboots when power returns.
There is no better way to guarantee that, than to yank the mains cable
from the UPS and see what happens (at a convenient time, that is).
>> When in doubt, it is always better to follow the instructions from the
>> genericups man page under the chapter 'TESTING COMPATIBILITY' to prevent
>> inadvertent shutdowns of your system.
> Yes, I guess you're right, but I start to get sick of it.
You have every reason to. You expected something to work (and if your
UPS indeed would have used the SafeNet software, it would, or else give
you an error message that the configuration was not OK) and instead it
is giving you all sorts of trouble. Please complain to the company that
sold you the UPS (or better yet, to Sweex for providing incorrect
information on their website).
> As I said before this is a live system and I can't mess with it to much.
> I really hope someone can give me the link I 'm missing here.
> Otherwise it's better to stop waisting my time on that ups and look for
> another one better supported. MGE?
If it should 'just work' and you don't find enjoyment in trying to make
it work, do yourself a favor and indeed swap it for one that is better
supported by the manufacturer (MGE indeed comes to mind, since they
officially sponsor NUT).
> Suggestions from recent models working without problems are welcome.
> I use the stable debian package, nut 2.0.1-4
That version is already quite old, we're at 2.0.4 already, with 2.0.5
lurking just around the corner.
More information about the Nut-upsuser