APC UPS WTFs



  • Last night, the battery in the UPS "protecting" my home server failed its self-test. This revealed a couple of WTFs.


    1. The battery self-test apparently involves temporarily disconnecting the UPS from the mains supply. This means that when the battery failed, the unit power-cycled the computer it was supposed to be protecting.


    2. When visiting the APC website to find out more, it turns out that "selecting a model" means something other than choosing an entry from the "select your model" dropdown menu:





  • Biggest WTF is the stupid cables APCs need so you have to spend something ridiculous when you buy one used without the cable.  RJ-45 *10* pin FTW!

     (Yes, I'm going to open the damn unit and just install the proper jacks that should have been there in the first place)



  • @shepd said:

    RJ-45 *10* pin FTW!

    That's 25% more lol in your cat 5, though!

     

    Some people will complain about anything... 



  • @Carnildo said:

    2) When visiting the APC website to find out more, it turns out that "selecting a model" means something other than choosing an entry from the "select your model" dropdown menu:
     

    Let me guess: you used your keyboard to select the model...



  •  Even better is their Windows drivers.Tell windows that you have an "APC Back-UPS" on an unconnected com port and it will repeatedly reboot shortly after booting until you either connect one or manage to tell it that actually you have an "APC Smart-UPS" before it reboots.



  • @Thief^ said:

    Even better is their Windows drivers.Tell windows that you have an "APC Back-UPS" on an unconnected com port and it will repeatedly reboot shortly after booting until you either connect one or manage to tell it that actually you have an "APC Smart-UPS" before it reboots.
    I'm confused.  Why does a UPS need a driver?



  • @belgariontheking said:

    I'm confused.  Why does a UPS need a driver?
    So that it can power down the machine cleanly before it's battery runs out.

    BTW, is it just me, or does PowerChute Plus forget the admin password after a few weeks? It's happened to me several times now, and apparently my coworker has the same experience (and the last time I installed it, I wrote down the password).



  • @Carnildo said:

    Last night, the battery in the UPS "protecting" my home server failed its self-test. This revealed a couple of WTFs.

    1) The battery self-test apparently involves temporarily disconnecting the UPS from the mains supply.

    Just exactly how good a test do you think it's going to be if it doesn't involve actually testing whether the dam' thing is going to work in a crisis or not?

    @Carnildo said:

    This means that when the battery failed, the unit power-cycled the computer it was supposed to be protecting.
    Yeh, right.  You should probably not be allowed near the "test" button on circuit breakers either...



  • @DaveK said:

    @Carnildo said:

    Last night, the battery in the UPS "protecting" my home server failed its self-test. This revealed a couple of WTFs.


    1. The battery self-test apparently involves temporarily disconnecting the UPS from the mains supply.

    Just exactly how good a test do you think it's going to be if it doesn't involve actually testing whether the dam' thing is going to work in a crisis or not?

    I expect it to run the test in a fail-safe manner, perhaps by connecting the inverter output across a high-wattage resistor for a few seconds.



    1. Usually, the battery will be detected as failing long before it can’t carry the load during the test. Did you perhaps under-rate the UPS for the connected devices. (I see you chose the 350 VA Back-UPS from the list; is the power supply in the server >= 200W1? Is there something else also connected?)
    2. To test with a resistor it would have to have something like this [url=http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=WFH230L75RJE-ND]resistor[/url] and a big hunk of aluminum for a heatsink.

    You should probably consider a higher-rated UPS and possibly one of the higher-end models, especially for something you’re classing as a “server”

    _____________________________
    1[url=http://www.apc.com/resource/include/techspec_index.cfm?base_sku=BE350G]Back-UPS 350ES Specifications[/url]



  • Connecting the inverter across a test load doesn't check failover, which is what's important. There are quite a few tests that it is generally accepted have to proceed in very specific ways in order to ensure that we can rely on their results.

    The test button a GCFI has to induce a small leakage current and success means the breaker trips. This means you can't test a GCFI without interrupting power. Yes, it's a bit of a pain. But it ensures that a successful test means it will really work.

    Now, the UPS test isn't trying to cause the computer to reboot if the battery was dead. That would be kind of pointless for an automatic test, but sensible for a manual test. It is generally accepted that a manual UPS test button can do nothing but disconnect the mains. Presumably, if you asked for the test, you want to see if it works or not -- all the way. Just like with a GCFI.

    You didn't say if this was a manual test you asked for or an automatic test it did on a schedule. If the former, you got correct behavior. If the latter, it's a bit of a WTF. An automatic test on a schedule should not significantly increase the risk of failure.



  • Most Back-UPS models don’t have manual tests, only the automatic one that runs after X-hours of on-time.I actually ran into the same problem at work; I had an ES 350 on my workstation, which was too small for the power supply.



  • @Carnildo said:

    I expect it to run the test in a fail-safe manner, perhaps by connecting the inverter output across a high-wattage resistor for a few seconds.
     

    I see. We've identified the actual WTF in this thread: that you don't understand what is required to test a UPS properly.

    Perhaps, instead of actually running a test next time, you can just flip a coin. You know, "Heads it passed, tails it failed.". Or wait - to be more like what you expect, it should probably be, "Heads it passed, tails it passed." 




  • @joelkatz said:

    The test button a GCFI has to induce a small leakage current

     It's GFCI - Ground Fault Circuit Interruptor


  • @ender said:

    BTW, is it just me, or does PowerChute Plus forget the admin password after a few weeks?

     If that software needs to store my admin password somewhere, I'm not letting it near my computer.


  • @alegr said:

    @ender said:

    BTW, is it just me, or does PowerChute Plus forget the admin password after a few weeks?

     If that software needs to store my admin password somewhere, I'm not letting it near my computer.

    Nothing like re-entering your admin password into a trojan every few weeks... Sounds intelligent to me!



  • @alegr said:

    @joelkatz said:

    The test button a GCFI has to induce a small leakage current

     It's GFCI - Ground Fault Circuit Interruptor

    You sure he wasn't talking about the Gulf and Caribbean Fisheries Institute?  Because that made a lot more sense than your whole "electricity theory".



  • @Sir Twist said:

    1. Usually, the battery will be detected as failing long before it
      can’t carry the load during the test. Did you perhaps under-rate the
      UPS for the connected devices. (I see you chose the 350 VA Back-UPS from the list; is the power supply in the server >= 200W1? Is there something else also connected?)
    2. To test with a resistor it would have to have something like this resistor and a big hunk of aluminum for a heatsink.

    You should probably consider a higher-rated UPS and possibly one of the higher-end models, especially for something you’re classing as a “server”

    1. I'm not sure what the total load on it was, but based on the runtime during the last power outage, I'd estimate it at 150 watts or so.
    2. Yes, that's about what I was picturing. I was expecting the bulk price to be closer to $10 than to $25.

    The server in question is a Pentium MMX with 48MB of RAM: more than powerful enough to run a webserver, mailserver, MySQL database backing the website and mailserver, caching DNS resolver, SSH server, FTP server, and assorted monitoring programs. It's probably not what you're picturing when you hear the word "server", though.


  • @Carnildo said:

    Last night, the battery in the UPS "protecting" my home server failed its self-test. This revealed a couple of WTFs.


    1. The battery self-test apparently involves temporarily disconnecting the UPS from the mains supply. This means that when the battery failed, the unit power-cycled the computer it was supposed to be protecting.

    Oddly enough, at work, they always schedule an 'outage window' whenever they do their periodic UPS testing - despite the fact that all of our gear is supposed to be dual powered.

    Every UPS I've ever personally used also disconnects power from the main when doing a manual test.  This includes APC, Belkin, and Tripp Lite, as well as a few whose brands I'm not recalling at the moment.  As such, it's more than just the other people on this board who think it's not a WTF that the self-test actually tests the UPS completely.


  • Garbage Person

    @tgape said:

    Oddly enough, at work, they always schedule an 'outage window' whenever they do their periodic UPS testing - despite the fact that all of our gear is supposed to be dual powered.

     

    Fool is the man who trusts that both power supplies are going to work (also, switches and such are usually also battery-backed and often lack dual power supply options)


  • @tgape said:

    Oddly enough, at work, they always schedule an 'outage window' whenever they do their periodic UPS testing - despite the fact that all of our gear is supposed to be dual powered.

    This is called 'being responsible'.



  • @tgape said:

    @Carnildo said:
    Last night, the battery in the UPS "protecting" my home server failed its self-test. This revealed a couple of WTFs.


    1. The battery self-test apparently involves temporarily disconnecting the UPS from the mains supply. This means that when the battery failed, the unit power-cycled the computer it was supposed to be protecting.

    Oddly enough, at work, they always schedule an 'outage window' whenever they do their periodic UPS testing - despite the fact that all of our gear is supposed to be dual powered.

    Every UPS I've ever personally used also disconnects power from the main when doing a manual test.  This includes APC, Belkin, and Tripp Lite, as well as a few whose brands I'm not recalling at the moment.  As such, it's more than just the other people on this board who think it's not a WTF that the self-test actually tests the UPS completely.

    This is the automatic every-fourteen-days test. Manual testing involves unplugging the unit from the wall, and in that case, if the battery were failing, I'd expect it to power-cycle the attached computer.


Log in to reply