UPS Issues

[a12vman]

I did a plugs-out test against my UPS last night.

 

I think I have something wrong with my UPS Settings.  

 

I have my Tripp-Lite 1500 setup(In Unraid) to stay on UPS Power for 5 Minutes. 

 

As soon as I went on UPS power unraid initiated a shutdown and was powered down within 45 seconds.

 

My server is the only thing plugged into the UPS. I wish it was an APC but seems to be recognized by UnRaid.

 

It's a Dell Server with redundant power supplies.  It appears to be consuming ~ 250 watts under normal conditions.

 

Is this a config issue or a problem with my UPS?  

Edited December 4, 2019 by a12vman

[S80_UK]

You appear to have nothing entered in the fields "Battery level to initiate shutdown" and "Runtime left to initiate shutdown".  From the details dispay it seems that these fields have therefore got some random data in them.  In particular the charge left to initiate shutdown seems to be at over 5400%, and therefore even a full battery (yours is at 100%) may not be enough to prevent an instant shutdown.  Try setting the first parameter to 50% and the second to maybe 5 or 10 minutes.

 

Also - a word of advice....  It is generally not recommended to unplug in order to test to test a UPS.  Instead, just switch off the power at the socket.  This keeps any earth connection intact.  Also, for testing, before cutting the power I stop the array and ensure all drives are at least spun up.  This means that if power to teh server then fails without a graceful shutdown, you won't later have to go through a lengthy parity check cycle.

[JonathanM]

28 minutes ago, S80_UK said:

Also - a word of advice....  It is generally not recommended to unplug in order to test to test a UPS.  Instead, just switch off the power at the socket.  This keeps any earth connection intact.

+1

I personally had a UPS fry some attached equipment (luckily just a test load) when I yanked the plug. Never again. The issue is if there is another path to ground somewhere, like through a network cable, it can suddenly be energized at full voltage when the UPS reference to ground is lost.

 

I recommend setting up a test rig specifically to deal with UPS troubleshooting.  You need...

2 Surge protectors with switches.

Dummy load roughly equivalent to the peak draw of your protected equipment. That can be incandescent lamps, small heater, hair dryer, or something similar.

Plug both surge protectors into the wall. Plug your server and all other loads to be protected into one of them that you keep turned on, and the UPS into the other. Plug your dummy load into the UPS. Boot up the server, and start all normal services.

Turn off the switch on the surge protector feeding the UPS.

Observe the behaviour of the server. Hopefully it shuts down properly well before your dummy load drains the UPS more than 50%.

 

It would be helpful if you had a kill-a-watt or similar meter to measure the actual draw of your equipment and dummy load.

 

Keep in mind that your network infrastructure also needs power during an outage so your notifications work properly.

[a12vman]

Wow you guys bring up some valid points that I never considered,  thank you for your input.

 

I will take the array off-line before doing a test like this again, and will turn off the breaker and keep the ups plugged in.

 

I cleared Time on Battery and entered values for %Battery Remaining for Shutdown. Now the attribute "MBATTCHG" shows 25 %.

 

I do like the idea of testing on a similar load. I measured 260 watts with all drives spinning and the server running under normal load.

 

I will find a similar load and test with that.

[a12vman]

I added parameters in UPS Settings for % Runtime remaing 25%.

 

I took the array offline and shut off the power at the breaker.

 

Checkng the status of UPS Settings from my laptop showed that the server went straight to shutdown.

 

I am trying to figure out whether my batteries are not holding a charge or I have some other issue. 

 

The attached log shows that the charge dropped to 68% but the server only ran on battery power for 60-90 seconds.

 

 

[JonathanM]

13 minutes ago, a12vman said:

The attached log shows that the charge dropped to 68% but the server only ran on battery power for 60-90 seconds.

300W load on that model shows 28 minutes of predicted runtime, so, yeah, either your batteries are bad or your load calc is wrong.

 

BTW, 90 minutes before time triggered shutdown is way too long. I'd leave that at 300 seconds. SLA batteries don't like deep discharge, I'd shoot to keep the batteries above 50% charge state at all times.

[a12vman]

I just did a power test with just a 520 watt heating element. 

 

The TrippLite ran for 6min 20sec with this load before shutting down.

 

Tripplite Spec sheet says I should get 15 minutes of run time with a 500 watt load.  

 

Edited December 8, 2019 by a12vman
Add'l Info

[JonathanM]

Good news, they look like standard batteries so should be available locally to you relatively cheaply, about $15-20USD

Bad news, you have to do all 3 at once.

 

[JonathanM]

BTW, before you do a battery swap, check to see if there is a learn function. Perhaps all you need to do is a full learn cycle to get things back in spec.