Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

At an Not For Profit/Charity I volunteer for occasionally we were very fortunate to be provided (donated) with a brand new 3Kva APC UPS. This is total overkill for our modest server rack (four mid-range servers and a switch!) but hey, I'll take what I can get!

Pressing the TEST button on the front indicates that yes, the UPS does work. Brilliant. But it only tests it for about 15 seconds.

My question is - will I degrade the UPS by unplugging it from the wall to see how long it will last? My plan is to unplug it, and wait until the battery meter reaches its last LED before plugging it back in, so that I know about how long I will have in the event of a power outage.

Do people do this on a regular basis? I'm guessing no (Lead Acid is very different to Li-Ion batteries)... but what kind of harm would it do if this were to happen (on purpose) every 6 months?

share|improve this question
    
3Kva is not overkill for four servers and a switch. Not if you want decent runtime before invoking the shutdown. –  kmarsh Aug 21 '09 at 16:12
1  
Please verify how many volts and amps the circuit this KVA is on. It should be 120V/30A though I suppose 208V/15A is a possibility. –  kmarsh Aug 21 '09 at 16:35
    
It's 240(230?)/15A - normal circuits in Australia are 240/10A. –  Mark Henderson Aug 21 '09 at 23:04
add comment

10 Answers 10

up vote 40 down vote accepted

You should consider turning off the circuit breaker to the outlet running the rack in lieu of unplugging the cord from the wall. The UPS is losing its electrical ground when you unplug it from the wall. While it's unlikely that anything would go wrong, the UPS designers "expect" that path to ground to remain available at all times, and if something did short during your test you might see sparks (smoke, flame, etc) when the electricity takes another path to ground. I've unplugged UPSs from the wall for testing before, but seeing a flash of "lightning" and hearing a loud "bang" coming out of a UPS during one such test gave me "religion" about not doing that again. After talking to an electrician friend I decided that, from then on, I'd do UPS tests that didn't interrupt the ground to the UPS.

BTW: The PowerChute Network Shutdown software from APC is garbage. You might have a look at apcupsd. It runs under a variety of operating systems (Windows included) and is much easier to configure (and to replicate the configuration on multiple servers via copying files) than the APC alternative.

share|improve this answer
3  
It's installed in a church building, so religion is not hard to come by there :) That said, I don't particularly like the sound of anything in your story so I'll keep that in mind. It's on a dedicated 15 amp circuit so killing the power shouldn't affect anything else. –  Mark Henderson Aug 21 '09 at 10:49
    
+1 that PCNS is garbage. Take a look at the image size when run on Sun's Linux 64bit Java. Pay attention to the defaults when using the automated install. After much effort I finally got APC to open bug reports on these and other issues, a year later gave up as they would do nothing about them. –  kmarsh Aug 21 '09 at 16:11
    
DANGER DANGER DANGER A 3Kva UPS should NOT be on a 15A circuit!!!! –  kmarsh Aug 21 '09 at 16:15
4  
@Cristian: If your equipment has a grounding prong on the plug then it needs to be plugged into a grounded outlet. If something shorts and there isn't a ground in the outlet then the electricity will find some non-preferred route to ground-- possibly through you when you touch the equipment (potentially making you dead). –  Evan Anderson Sep 1 '09 at 12:04
1  
Every socket in Australia has a ground, thankfully. –  Mark Henderson Feb 10 '10 at 18:59
show 4 more comments

Read the manual. :) (Though that can be a bit difficult to find on APC's page...)One thing I've not seen mentioned anywhere is that there is a user-initiated "Runtime Calibration" option on APC's stuff. You can do it either via powerchute software (I recommend the "Business" edition) or via some combination of the test/power buttons on the front. As the documentation states, you want to do with a similar but non-important load.

share|improve this answer
    
APC calls this process the ''Brain-Dead'' process. Here is documentation from APC on how to do this for a Back-UPS and for a Smart-UPS. –  Mei Oct 31 '11 at 14:57
add comment

"indicate you're getting a lot of over volts" - APC UPS sometime come with low max voltage settings which will cause lots of overvoltage events. Raise the limits so they are in line with Australian spec. Many of the units I looks after in the country frequently see variations in voltage and many are on the high and of 250 volts..

share|improve this answer
add comment

I'm going to make this point loud and clear.

DO NOT TEST THE UPS BY UNPLUGGING IT

You break the ground, which means if any of the hardware has a short, and there is no other ground it will go though you to get to ground if you end up being the shortest path. If everything is working fine this will never happen, but hey if this never happened there'd be no need for a ground.

Best way, is to have the outlet that the UPS is plugged into able to be switched on and off so that the gound and netural will remain intact though the test. The breaker can do this or you can do local high quality switch. If you can't do that, then put a VERY GOOD (ie $30 to $60 range) power bar with an off switch between the wall and the UPS, make sure you label the switch for what it is for. The point in the other post that is mentioned in the comment is to NOT over load the powerbar, doing it this way is better then unplugging it. You can now switch off the line in and simulate a power failure, this will leave the ground and netural intact.

You can test by letting it run down, although crude it will work. Also if the software has a calibration option it will do that for you and run it ever 6 months or so. The run time will degrade over time, so if you are using monitoring software to shut down the servers at say 15%, that 15% will change over time and it can correct for that.

For your voltage issue, if you can run a decided line from the power box to the servers and us it only for the UPS. Things like tube lights, fans, motors, etc will create dirty power. Having the server on its own will help that since it moves it to a circut further away for those items. If its still happening it might be worth getting an isolator put in or it could be that your utility power is just really bad. Put a good meter on to a line and see what its really reading. It needs to be a good meter because I've seen cheap ones be off by 5 volts and thats enough to cause a UPS to go into over voltage so you need an accurate number. If this is a church, there is a chance you have an electrican as a member that could help out.

Here are reference links to grounding and daisy chaining UPS

Grounding

Daisy Chaining

share|improve this answer
1  
Plugging a UPS into a power strip is apparently a Very Bad Idea. serverfault.com/questions/29288/… –  Kara Marfia Aug 22 '09 at 0:18
2  
It's good to understand why its a bad idea. The reason it's a bad idea is because people over load the plugs by thinking adding 20 plugs allows them to plug more stuff in when the line has a finite limit. In this case it's better to use a power bar with a switch then to unplug the UPS and the plug is also on a dedicated line. The best way is to make it a switched outlet that the UPS is plugged into though. I updated my answer to reflect that and added some links. –  SpaceManSpiff Aug 22 '09 at 9:58
    
Unfortunately, those links are dead now. –  Mei Oct 31 '11 at 14:50
add comment

Testing the batteries to the last led equates to roughly 80% Depth of Discharge. Doing this test on a regular basis will reduce the life of the batteries much more than a more shallow depth of discharge. I would recommend only testing to 50% Depth of Discharge.

Battery life is directly related to how deep the battery is cycled each time. If a battery is discharged to 50% every day, it will last about twice as long as if it is cycled to 80% DOD. If cycled only 10% DOD, it will last about 5 times as long as one cycled to 50%. Obviously, there are some practical limitations on this - you don't usually want to have a 5 ton pile of batteries sitting there just to reduce the DOD. The most practical number to use is 50% DOD on a regular basis. This does NOT mean you cannot go to 80% once in a while. It's just that when designing a system when you have some idea of the loads, you should figure on an average DOD of around 50% for the best storage vs cost factor. Also, there is an upper limit - a battery that is continually cycled 5% or less will usually not last as long as one cycled down 10%. This happens because at very shallow cycles, the Lead Dioxide tends to build up in clumps on the the positive plates rather in an even film. The graph above shows how lifespan is affected by depth of discharge. The chart is for a Concorde Lifeline battery, but all lead-acid batteries will be similar in the shape of the curve, although the number of cycles will vary.

http://www.windsun.com/Batteries/Battery%5FFAQ.htm#Lifespan%20of%20Batteries

share|improve this answer
add comment

Once you have the PowerChute software installed and running, you can also install additional clients thereof on the other servers and configure the whole lot for automatic shutdown once the battery level drops below a certain percentage, triggered by the one server that has the USB/serial connection. The newer versions of UPS will also support this via Ethernet, but this will require an additional plug-in card. This can be very helpful, especially with database servers (which don't like hard power failures at all).

share|improve this answer
add comment

If it's a 3Kva APC UPS chances are real good it has either a serial or USB port on the back (hope for the USB). IF that's the case, you can connect a Windows or Linux box to it and run PowerChute (available for download at apcc.com). It should tell you the expected run-time of the UPS itself. Since it sounds like you have a light load, it may be pretty long ;).

However, upthread you indicate you're getting a lot of overvolts. This will unfortunately reduce the lifetime of the UPS itself as it'll be dealing with all that dirty power. Perhaps once or twice a year run an unplug-from-wall test, and watch the run-time level in powerchute to see if it is still accurate. If it starts decrementing a minute every 30 seconds, which I've seen happen, you know your runtime estimates are buggered and it's time to retune your shutdown procedures. And it's time to get new batteries.

share|improve this answer
add comment

First off, that most certainly is not overkill. It's a reasonable size for what you have. As for the testing, if you have Windows install the accompanying software (or download it from the APC site) and use it to perform the testing and calibration of your UPS.

It's worth mentioning that, depending on model, an APC UPS will normally self test every week or month, which includes running down the batteries to determine health and run time.

You might also consider reducing the sensitivity setting of the UPS if it is frequently compensating.

share|improve this answer
1  
I though it might be overkill because only one the little LEDs that indicate load lights up, and only after I power on the last server (before that, it's got 0 lit up). I haven't done any configuration of it yet, so I'll take a look at the sensitivity settings on it. –  Mark Henderson Aug 21 '09 at 5:18
1  
One light is nice. That means you have a good run time and the UPS should never get stressed. :) –  John Gardeniers Aug 21 '09 at 5:33
add comment

Testing with an unplugged UPS is very similar to testing a real power failure.
This would add a nearly-full recharge-cycle to your batteries.
You are correct about the difference with Li-ion batteries.

If you have a software link with the UPS, it will trigger alarms and eventually a shutdown.
You could re-plug on the first alarm.

Caveat: Remember that after such a test you are left with no charge on the batteries
a real power failure at this time would leave you with no backup power.
And, Lead-Acid batteries take longer to recharge -- extending your critical no-backup window.

You can read up more on Lead-Acid batteries at the Battery University page.

It takes about 5 times as long to recharge a lead-acid battery to the same level as it does to discharge.
On nickel-based batteries, this ratio is 1:1, and roughly 1:2 on lithium-ion.

share|improve this answer
    
Good point - don't want a real power failure after the test! I notice that it's almost always compensating for over-voltage as well, which kind of indicates the quality of the power in the building... –  Mark Henderson Aug 21 '09 at 4:32
add comment

No, it will not degrade it. Feel free to test - just be careful since you may not have enuf time to do a clean shutdown. Computers/Servers use more power when booting up and shutting down compared to just sitting idle.

Another way is if you can determine what power your server rack draws and creating a similar load and see how long it stays up. But its never a good idea to use 100% of the juice from batteries since you can never tell how much juice they actually have. I'd be careful going over 75% of the total time.

If you need the maximum uptime, i'd look into powering down unwanted servers and letting just hte critical servers run..

share|improve this answer
    
For a first test I'd just plug in a couple 100 watt light bulbs. –  chris Aug 21 '09 at 12:08
    
Hehe incandecent bulbs are illegal in Australia now. You can still use them but you can't buy them. You can only get the fluro energy efficient ones, and I think you'd need a lot of them to get any kind of load! –  Mark Henderson Aug 21 '09 at 23:29
    
haha come on guys, there's always other options: Workstations, heaters (u can control the temp and thus sort of control the flow of current), hair dryers?, etc.. Im sure you can come up with many ways to burn electricity =) –  J Sidhu Aug 22 '09 at 5:13
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.