Servers tripping GFCI

Users who are viewing this thread

Beekerc

IT Consultant / Network Engineer
Messages
94
Reaction score
0
Points
0
Location
Seattle
I'm currently remodling my basement. As such, I have had to move my rack of servers (4 dell servers, 2 pc's, 2 APC SmartUPS 1000VA UPS'es) out into the garage, until my basement office is finished. I have a 20A circuit servicing the garage, with a faceless GFCI (you may recall this from another of my posts) between the panel and all the outlet runs. Originally this circuit was feeding outlets inside the basement, so no GFCI was requried, but once the carport was enclosed, I simply flipped the outlet boxes in the stud bay to point into the garage, hence the need for GFCI protection - but while I was waiting for them to arrive, I left the circuit live, commenting to the inspector that this was technically "old-work" and that the GFCI that I wanted was on order and would be in place by final (the outlet box was properly wired for it) and he said this was fine. The The problem is that before the I placed the GFCI in, power was running fine, voltage levels were fine and the outlet tester reported correct wiring. I outlet test *every* outlet that I install. Once I put the GFCI in, it would trip within 5 to 10 minutes, after power-up and reset.

Since I actually have two 20A circiuts in the garage, i ordered two faceless GFCI's. thinking it could be a faulty unit, I tried the other GFCI, and even tried a 20A GFCI outlet (to be installed in a bathroom) to see if it were a faulty protector. every unit tripped withing 5 to 10 mintues.

I slapped a clamp-on meter to the wire and saw that it was drawing about 14 amps - well within the tolerance of a 20A circuit on 12/2 wire.

I again checked all the receptacles for wiring faults and voltage drop and everything looked fine.

given what i've read about nuisance tripping, especially with regard to motors, is there something to be said about servers, UPS's or surge protection strips on GFCI protected circuits, that may cause them to trip? or is there a deeper problem here?

I suppose the computers could be considered a "permanent appliance" in which case they would not need to be GFCI protected, despite the location, but the repeated tripping is anomalous behavior that has me searching for an explanation.

Thanks
BeekerC

PS. please, no lectures about how I need have the GFCI installed regardless.
I am aware of the risks, and I have mitigated them as such.
1) currently on the back of the garage is actually wired (live outlets are at least 15 feet from any exterior door - ie. water source)
2) only one outlet is being used, by the servers, all others are capped with child proof plugs and will not be used until the GFCI is installled
3) once the servers can be moved into the basement, the GFCI device will be installed.

in my limited, but semi-educated, opinion, this outlet poses no greater risk than the originally installed, non-GFCI outlets on my bathroom countertops - which have not managed to shock or kill in 18 years and will be upgraded to GFCIs later this month.
 

JWelectric

Electrical Contractor/Instructor
Messages
2,608
Reaction score
21
Points
38
Location
North Carolina
If the GFCI is not tripping with nothing connected to the circuit and then the servers connected they trip should be self explanatory.

Try connecting something else into the receptacle for an extended time and see if it still trips. If it does it is possible that there is a damaged cable in the wall that is impeding current flow.

If not then we will know that the UPS or surge protector for the servers is letting current flow across to eliminate spikes from the servers.

If the latter is the case then just remove the GFCI device until they can be moved as I don’t see another solution.
 

Beekerc

IT Consultant / Network Engineer
Messages
94
Reaction score
0
Points
0
Location
Seattle
If the GFCI is not tripping with nothing connected to the circuit and then the servers connected they trip should be self explanatory.

agreed, just wondering if there might be an explanation as to why.

Try connecting something else into the receptacle for an extended time and see if it still trips. If it does it is possible that there is a damaged cable in the wall that is impeding current flow.

havent' tried a different load (alone) for an extended period, but that would be a good test.

If not then we will know that the UPS or surge protector for the servers is letting current flow across to eliminate spikes from the servers.

agreed. was hoping someone might know if UPS or surge protector behavior might typically be a cause for nuisance tripping.

If the latter is the case then just remove the GFCI device until they can be moved as I don’t see another solution.

yup, that's the plan. will defintely try the alternate load before the final trim-out.

Thanks
 

Thatguy

Homeowner
Messages
1,454
Reaction score
3
Points
0
Location
MD
it would trip within 5 to 10 minutes, after power-up and reset.
UL says, for trip times for GFIs, T in seconds = (20/I)^1.43, with I in mA. This is a design goal, they are actually trip quicker.
So, I = 20/(T^0.7).
For T = 300 seconds, I = 0.4 mA.
Maybe 90' of Romex downstream of the GFI could have this much reactive current, but I think GFIs are supposed to have a cutoff so they don't trip on currents this low, ever.

This one
http://www.national.com/ds/LM/LM1851.pdf
has a minimum trip threshold of 3 mA differential current, but Note 3 on this datasheet says adjustment of the sensitivity may be necessary.

So, not-properly-adjusted-at-the-factory GFI?

If you're comfy with electricity, you can check downstream cable leakage current with a voltmeter, a 10kΩ, 2w resistor and some clip leads.
 
Last edited:

Beekerc

IT Consultant / Network Engineer
Messages
94
Reaction score
0
Points
0
Location
Seattle
UL says, for trip times for GFIs, T in seconds = (20/I)^1.43, with I in mA. This is a design goal, they are actually trip quicker.
So, I = 20/(T^0.7).
For T = 300 seconds, I = 0.4 mA.
Maybe 90' of Romex downstream of the GFI could have this much reactive current, but I think GFIs are supposed to have a cutoff so they don't trip on currents this low, ever.

panel to server outlet is 25 feet, 30 feet max.
I don't follow all that math, but yes, my understanding is that if a fault is detected, it's supposed to trip instantly - as witnessed by the GFCI outlet tester I use, push the button, the GFCI (wherever on the circuit) trips.

but Note 3 on this datasheet says adjustment of the sensitivity may be necessary. So, not-properly-adjusted-at-the-factory GFI?

a thought, but i tried two other GFCI's, one identical faceless (by Hubbell) and one receptacle (by Leviton). all three behaved the same way, so that would rule out faulty or improperly adjusted equipment.

If you're comfy with electricity, you can check downstream cable leakage current with a voltmeter, a 10kΩ, 2w resistor and some clip leads.

I'm comfortable working with electricty, but i'm not sure what you said here, much less sure of how to rig up that resistor to test for leakage.
 

Billy_Bob

In the Trades
Messages
419
Reaction score
0
Points
0
Some surge protector power strips have a little LED between ground and hot to indicate a good ground and/or some minimal circuitry between ground and hot.

So might try removing any power strips/surge protectors you have which have LED's on them and see if that solves the problem.

(Of course a current to ground would trip a GFCI, although an LED would be a quite small current, thus the not tripping right away.)
 

Thatguy

Homeowner
Messages
1,454
Reaction score
3
Points
0
Location
MD
panel to server outlet is 25 feet, 30 feet max.
I don't follow all that math, but yes, my understanding is that if a fault is detected, it's supposed to trip instantly - as witnessed by the GFCI outlet tester I use, push the button, the GFCI (wherever on the circuit) trips.
The test button puts ~8mA differential current so it is sure to trip.

a thought, but i tried two other GFCI's, one identical faceless (by Hubbell) and one receptacle (by Leviton). all three behaved the same way, so that would rule out faulty or improperly adjusted equipment.
I agree that is not likely to be the GFI.


I'm comfortable working with electricty, but i'm not sure what you said here, much less sure of how to rig up that resistor to test for leakage.

With everything unplugged, you inject 120v with respect to ground into the hot lead of the unpowered (H & N disconnected at the testing end) cable through a series resistor.
With 10k, 4 mA leakage (H to G in the cable under test) should read 40v across the resistor on a voltmeter set to the 120v scale.
With your 25' Romex setup, less than ~0.1v across the resistor passes.

You can also use just a milliammeter to read the 4mA, but if a short suddenly develops, you pop your meter's internal fuse (been there, done that). The resistor setup can handle a short but it will warm up, and digital meters have pretty good overvoltage protection in case the resistor voltage goes to 120v.

BTW, LEDs take at least 10 mA.
 
Last edited:

Billy_Bob

In the Trades
Messages
419
Reaction score
0
Points
0
P.S. If you have many power strips, you might be able to track down the offending strip(s) by plugging one in using a 3 prong to 2 prong plug adapter (for plugging in 3 prong plugs into 2 prong outlets).

Then see if one of the LED lights on the power strip no longer works with the adapter. If yes, then this power strip is using a little electricity to ground to power the LED.
 

Beekerc

IT Consultant / Network Engineer
Messages
94
Reaction score
0
Points
0
Location
Seattle
P.S. If you have many power strips, you might be able to track down the offending strip(s) by plugging one in using a 3 prong to 2 prong plug adapter (for plugging in 3 prong plugs into 2 prong outlets).

Then see if one of the LED lights on the power strip no longer works with the adapter. If yes, then this power strip is using a little electricity to ground to power the LED.

I don't recall the brand, but they are high end surge supressors, either APC or Belkin. i do know they have a red and green LED and the "leakage to ground to power the LED" makes sense. I'm not sure i'd want to run any computer equipment on a non-surge protected line, nor do i have a way to octopus many plugs to one receptacle without some sort of surge device. I'd try the 3-prong/2-prong test if i owned such an adapter - there are no 2 prong receptacles in my house so never had a need for such an adapter.

I think i'll go with Occam's Razor on this one in that the simplest explanation is probably the right one. Esp since i have a work around which is to leave the GFCI's out until I move the servers to their new home.

to ThatGuy: thanks for explaining the testing, but as soon as you mention "injecting" voltage and burning out a meter, i pretty much lost my enthusiasm for that line of analysis. Appreciate the info though.
 

Thatguy

Homeowner
Messages
1,454
Reaction score
3
Points
0
Location
MD
to ThatGuy: thanks for explaining the testing, but as soon as you mention "injecting" voltage and burning out a meter, i pretty much lost my enthusiasm for that line of analysis. Appreciate the info though.
In that case I won't tell you some of the stuff that I got away with.

You can measure just the resistance from H to G in the unpowered cable. If it's less than 30k you've found the problem.
 

Billy_Bob

In the Trades
Messages
419
Reaction score
0
Points
0
I don't recall the brand, but they are high end surge supressors, either APC or Belkin. i do know they have a red and green LED and the "leakage to ground to power the LED" makes sense...

Yes it would be the higher end surge suppressors that have this "feature".
 

Beekerc

IT Consultant / Network Engineer
Messages
94
Reaction score
0
Points
0
Location
Seattle
Yes it would be the higher end surge suppressors that have this "feature".

it's a belkin. there is a red for grounded, green for protected and a light for the rocker switch. they all look like those wheatgrain neon lights, but they could be LED's. either way, my guess is the consensus is right and it's leaking return voltage down the ground to power the indicator lights
 

Fidodie

New Member
Messages
94
Reaction score
0
Points
0
Location
new jersey
if there is an led to tell you the ground is good - and it isn't something that requires you to press a switch, I'd say you found the source.

on another note, "critical" equipment, even in a wet environment, should not be on a gfci - consider a sump pump or sewage ejector...you wouldn't want a nearby lightning strike tripping off your pump just when you need it most!
same would happen to the servers - you might be away - and then run out of back-up with no way to reset.

I'll let you make the argument about what is critical. there is probably fault protection in the UPS just in case it was to get wet.
 

Beekerc

IT Consultant / Network Engineer
Messages
94
Reaction score
0
Points
0
Location
Seattle
same would happen to the servers - you might be away - and then run out of back-up with no way to reset.

aside from the fact that the chances of my servers getting wet are extremely low and that once inside the basement i wasn't planning on having them on a GFCI circuit anyway, your point is probably the best argument for leaving them off GFCI for now. I was even thinking about getting a network based PDU so I could do remote shutdowns and reboots, but even that would have been useless behind a tripped GFCI. I guess since the wet/shock risk was minimal and the wiring was correct, I hadn't even thought about not being able to reset remotely.
 

Brownizs

In the Trades
Messages
193
Reaction score
0
Points
0
Location
Springfield, IL
If setting the servers up in a basement, you should have a water alarm to warn if water intrudes on the space that they are in, and also, as mentioned before, you should not have them hooked up to a GFCI, but should have all mission critical equipment hooked up in a way for redundancy, and also as above mentioned, an alarm to warn of problems.
 

Thatguy

Homeowner
Messages
1,454
Reaction score
3
Points
0
Location
MD
Computers pull current with a non-sinusoidal waveform; they are more like pulses, which makes the power factor not equal to 1.
Maybe someone can say how GFIs will act with this kind of current.
 

Jadnashua

Retired Defense Industry Engineer xxx
Messages
32,771
Reaction score
1,191
Points
113
Location
New England
Some UPS devices power the connected devices entirely off of conditioned power...in other words, the line voltage keeps the batteries charged, but the thing is always running off of the battery-powered oscillator generated sine-wave. It would depend on the type of power supply they had on the charger as to how much noise it generated on the prime power line - a switching power supply would put a lot of noise on the incoming power line, probably as bad as a computer, but it may not be that type of ps charging the batteries.
 

Beekerc

IT Consultant / Network Engineer
Messages
94
Reaction score
0
Points
0
Location
Seattle
If setting the servers up in a basement, you should have a water alarm to warn if water intrudes on the space that they are in, and also, as mentioned before, you should not have them hooked up to a GFCI, but should have all mission critical equipment hooked up in a way for redundancy, and also as above mentioned, an alarm to warn of problems.

good advice, esp since the new server closet (a real closet for sound isolatiion) borders the same room as the washing machine. however, the basement drain is actually right below the dryer and these are 15 feet away from the servers. a water alarm may be inexpensive piece (peace?) of mind.

the only reason for having the GFCI's in the first place is because this is in the garage, and i wanted to get them wired in and be done with it. given this sequence of events, looks like the final trim out of these devices will have to wait.
 

Drick

In the Trades
Messages
460
Reaction score
19
Points
18
GFIs don't just sense power returning on the ground. They can sense the power going out on the hot leg and the power returning on the neutral leg. An imbalance caused by computer circuitry (fluorescent lighting is also known to do this) will cause the GFI to think there is leakage from the hot leg to ground somewhere and cause it to trip.

Without this hot/neutral power sensing ability it would defeat the purpose of installing GFIs in the kitchen and bath. How many hair dryers and mixers, coffee makers and crock pots have you seen with a grounded cord?
 
Last edited:
Top
Hey, wait a minute.

This is awkward, but...

It looks like you're using an ad blocker. We get it, but (1) terrylove.com can't live without ads, and (2) ad blockers can cause issues with videos and comments. If you'd like to support the site, please allow ads.

If any particular ad is your REASON for blocking ads, please let us know. We might be able to do something about it. Thanks.
I've Disabled AdBlock    No Thanks