[Nut-upsuser] false alerts/shutdown

Arjen de Korte nut+users at de-korte.org
Fri Jun 15 10:19:50 UTC 2007


>> What makes you think these are false alarms?
> I have 2 reasons for believing these are false alarms:
>
> 1) The UPS doesn't beep and none of the status leds change when this
> occurs. I configured nut to page me for 'on battery' events, and while
> working in the server room I witnessed how NUT complained but the UPS
> didn't give a sign. In contrast, when it goes on battery because I pull
> the plug, it beeps, leds flash, ...

How long it takes before a UPS starts to beep and whether or not it will
beep at all for glitches, is something that is rarely specified. Don't
count on it.

> 2) Three other servers (not backed up by ups) who share the same power
> distribution unit as the UPS are not affected. If these alarms are real,
> it means voltage drops to 160V. I thought that would shut a server
> down.  Maybe you are right and a server can handle this for a few seconds?

Not necessarily. When a UPS detects a mains glitch, it will usually go to
battery for a couple of seconds and then return. So even if the power
fails for just a couple of milliseconds, it will be on battery for a few
seconds. Most (if not all) servers will tolerate a complete loss of input
power for at least a full cycle (20 ms) and even longer than that for
brown outs. So the fact that other systems don't notice this, is no
indication either.

[...]

> I'm almost sure the mains are OK since the building is less than 2 years
> old.

That by itself tells nothing. Remember the recent deaths of patients in an
italian hospital where the builder swapped nitrous oxide and oxygen lines?
That department was brand new too, but sadly eight people died before they
found out.

> Also I connected the UPS to another socket across the room to no
> avail. I agree there might be electrical glitches, but wouldn't the UPS
> beep then?

Not necessarily. Many will only start to beep when the mains is really
out, to prevent nagging people for each and every transient.

To rule out problems with the USB port, could you try connecting this UPS
via a serial port and the 'mge-shut' driver? You might also run it in
debug mode, but this will probably result in huge amounts of data since
the protocol this UPS uses is very verbose.

Best regards, Arjen




More information about the Nut-upsuser mailing list