[rescue] It lives!

Skeezics Boondoggle skeezics at q7.com
Mon Jan 27 15:37:10 CST 2003

On Mon, 27 Jan 2003, Dave McGuire wrote:
>    Ok this is just excessive.  But I love excess.  Do you have an  
> ammeter handy?  I'd love to know what that box pulls.

Well, funny you should ask, because I was interested in that too. :-)  I
had the machine plugged into a 6KVA Best Power Axxium Pro which is idle at
the moment, so the 2000E was the only load on it.  It's wired through a
transfer switch for 208V in and out.[1]  According the output status page
provided by the web/SNMP card:

	Voltage Out (L to N) (VAC)	120.0/120.0
	Current Out (L) (AC Amps)	8.0/8.9
	True Power (L) (KWatts)		0.98/1.04
	Apparent Power (L) (KVA)	0.96/1.06
	Power Factor (L)		0.97/0.67
	UPS Load (L) (%)		32/33

Which was curious, since the numbers on the front panel display were
roughly half that - but the panel displays both the "total" and the
individual phases!  It didn't make sense to me that the Iout "total" would
be 4.4 amps (at 208!?), but the Iout L1-N and Iout L2-N readings were both
4.n amps (at 120?!?) as well!  I should bug Best/Invensys/Powerware about
that and see why the monitoring card and the front panel don't agree...

But from those numbers, would I add the watts for each phase for a total
of ~ 2.02KW?  Given that disk activity was causing the load to fluctuate
while Jumpstart was doing its thing, all these numbers seem a little bit
off.  16.9A @ 120V = 2028VA, but the power factor is all wonky, and I
can't figure how they get those watts numbers.  (Note that without the
SSAs powered on, the machine runs at a power factor of .98-.99 on both
phases.  This is the cabinet with four smaller blowers, not the one big
honkin' unit in the older 2000's.)

Anyway, the more I think about all that the more my head hurts. :-)

Next time I can roll the beastie down the hallway and park it outside the
machine room and play some more, I'll take more readings and try to make
some sense of them.  For now, "about 2000 watts" seems a reasonable
estimate... the SC2000 (non-E) I measured the same way (four system
boards, no SCSI trays) pulled about 700W... 

-- Chris

[1] Okay, time for An Amusing Little Story about Mr. Sparky.

This UPS was a lot of Fun to wire up.  The building provides three phase
in a typical 120/208-Wye (but we couldn't afford a nice beefy three phase
unit).  Our UPS takes 200-240V single phase and most of our loads are
plain old 120V.  So we pull 208V single phase off the three phase main
panel, through a make-before-break type transfer switch, putting 208V out
through a subpanel to PDUs that split out 120V plugs for the servers.  We
have two of these units, configured for 9KVA each, wired identically (or
so we thought).  The first one came up just perfectly.

The second one was ready for testing at 3:30PM on a Friday, as the crews
were finishing construction and everyone was cleaning up from a very busy
week (new A/C unit, new flooring, new racks, new power panels - a huge
undertaking).  We were ready to test out the transfer switch and bring the
2nd UPS on-line.  Service mode okay, pass-through mode okay, Line mode
<BANG>.  Very Bad Noise.  Bright Flashy Light.  Magic smoke escaping.  
Nasty burning electrical smell.

The look on the electrician's face - and everyone else's in the room - was

So after resetting not only the 60A input breaker, but the 100A main on
the building panel too (yikes :-) we pulled everything apart again,
checked all our connections, metered out all the voltages as laid out in
the startup checklist, everything is fine.  So we figure, okay, bad power
module, let's pull out the one that smoked, try it again.  It's Friday,
after all, and mostly we just want to go have Beers.  (Lots of beers, at
this point.)


Attempt #2 fails, slighly less spectacularly, but now there's this sense
of desperation and a hint of sheer terror.  We power it all down and call
it a day.  At least I know from the torrent of messages to my cell phone
that my SNMP trap receiver and paging system is working, so it isn't a
total loss, eh?

Call in the service guy.  A week later, with a pile of replacement parts
in hand, he arrives with manuals and meters and all kinds a things, does a 
thorough once over, checks everything out, tests the transfer switch.



Okay, so now I'm concerned.

So what did it turn out to be?  Well, the documentation said that if there
was a phase rotation issue, the machine would just say so on the front
panel and refuse to power up.  Now, you and I might see a slight
discrepancy between "display a message on the LCD panel" or "practically
explode and freak everyone out" and refuse to power up, but apparently
their technical writers didn't think so.  You must admit, we definitely
got the message...

Ultimately we found that L1 & L2 had been reversed in one of the four sets
of connections (main -> switch -> ups-> switch -> output panel), which is
what resulted in the light show.  Ironically, I had *specifically* asked
the electrician about following L1 & L2 through the entire circuit for
precisely that reason... seems the UPS didn't share the electrician's view
that "it's single phase, a hot's a hot."  I'm just an amateur, y'know?  I
love doing wiring projects - haven't burned anything down yet, and I like
to think I do very clean, professional work.  But I don't have much
experience with three phase applications, so I took his word for it... Ah,
well.  Hooray for warranties!

And remember kids, always respect Mr. Sparky. :-)

More information about the rescue mailing list