BUILD UPS tip

[Sir.Robin]

Exactly. And you can also adjust the timer for UPS shutdown. 5min are default.
If it works, i have yet to test.

This is on a PureSine series. Dunno bout the others.

 

[no_connection]

pull power from it so it will start??? Define "it" please.

"It" would here be the server as most can be set to start when power is applied regardless of how or if it was shut down.

Meaning if the UPS pulled power and then applied it again the server would switch on if it was previously shut down after a certain time on battery.

It would not make sense for a UPS to have to be manually started however.

Some can be set o require a certain amount of runtime before it applies power.

I have 30min + runtime out of mine so I manually manage it. Mostly I pull the plug when a thunderstorm is near just in case, so I can leave stuff running in the meantime.

A separate small UPS handles network for as long as it can.

 

[jgreco]

It might actually make perfect sense for a UPS to have to be manually restarted. It really depends on the target environment. An inexpensive UPS sold as protection for desktop PC's, a market segment known to be targeted by UPS manufacturers, could reasonably keep the load powered off until turned on by the user. If a PC is attached and deemed to be in charge of controlling the UPS, powering on without that PC *also* powering on would mean that nothing was managing the UPS. Since many PC's do not auto-power-on, that could be a problem... especially for the next power loss event that happens 15 minutes later, at which point the UPS drains itself trying to maintain a nonexistent load, waiting to coordinate a shutdown event that will never happen because the PC never auto-powered-on.

 

[no_connection]

I get your point. If the UPS does not have enough runtime left when a second power outage hits, meaning the server starts halfway and never have time to gracefully shut down when the UPS runs out again it would have been better to manually start it. But then you have to be in place to do that.
As you said, it depends on your intended usage and specific demands.

 

[jgreco]

No, that wasn't my point at all. My point was that if the use case for the UPS was intended to be a PC, then it is likely to have a different set of design goals. A server could be reasonably be something you would want to power back up automatically. A PC probably isn't. These units don't restart at all, or so it sounds like.

My guess is that the design goal for the UPS series being discussed is "desktop PC." So while I fully appreciate that most FreeNAS hackers want to grab the cheapest device that they can to do the job, my experience is telling me that UPS units intended to support servers, like the APC Smart-UPS series, do in fact have configurable options for restarting, including options to wait for the UPS to be recharged to a certain level before re-energizing the load.

So maybe the takeaway here is that if you buy a UPS intended for a desktop PC, make sure it has restart characteristics you are happy with. 'Cuz it probably doesn't.

 

[cyberjock]

All of the UPSes I own are more than 2 years old, and none of them turn off automatically under any circumstances. They run until power is restored or the battery is dead. Only the newer ones seem to automatically turn off.

 

[jgreco]

All of the UPSes I own are more than 2 years old, and none of them turn off automatically under any circumstances. They run until power is restored or the battery is dead. Only the newer ones seem to automatically turn off.

I don't really know about the Cyberpower stuff, but the APC's certainly can and do perform coordinated intelligent shutdown and even restarts. I pretty much got rid of all our pre-1997 gear some years ago so I can't go back further than that to check...

Code:

> 4
 
------- About Smart-UPS 1400 named apc12    -----------------------------------
 
        Serial Number    : WS9740226837    Firmware Revision : 70.9.D
        Manufacture Date  : 10/01/97
 
        Press <ENTER> to continue...

That's the oldest unit I could find still energized and on the network.

Code:

> 3
------- Shutdown Parameters ---------------------------------------------------
        Shutdown Delay       : 020 sec     Return Delay    : 000 sec
        Low-Battery Duration : 002 min     Sleep Time      : 0.0 hrs
        Return Batt Capacity : 000 %
     1- Shutdown Delay       (s): 020
     2- Return Delay         (s): 000
     3- Low-Battery Duration (m): 02
     4- Sleep Time           (h): 0.0
     5- Return Batt Capacity (%): 00
     6- Accept Changes          :
     ?- Help, <ESC>- Back, <ENTER>- Refresh, <CTRL-L>- Event Log
>

Now since this unit's under the shop bench there's no actual managed shutdown/startup configuration set up here.

But basically if it is talking to a UPS client on an attached PC, when the agent determines that it is time to shut down the PC, it is supposed to halt the PC and then signals the APC, which waits "shutdown delay" and then drops the load. It will also forcibly drop the load at the low battery threshold, warning the PC agent about "low battery duration" minutes before doing so. It can also be set to restart when the battery has recharged to a certain percentage.

These are not new UPS features. They did, however, used to be "high end" features and I'm guessing that there's less reason to differentiate these days, so the featureset on lower end devices is gaining some of the features that it arguably should always have had.

Of course, if the device can't be configured to behave in the manner you want, then that's bad.

 

[DrumNBisco]

As someone who, up until a year ago, used to work for APC, believe me when I tell you that there are more differences than just price when comparing their Back-up and Smart-ups product lines. I did tech support for pretty much every product they made except cooling and UPS' above 16 kVA for years,

The main difference would be inverter components and sine-wave output. The back-ups outputs a stepped sine-wave when on battery, while the smart-ups output a pure sine-wave while on battery. This is fine for most power supplies (it will not work for motors eg aquarium pumps), however I have seen issues with back-ups and active power factor correcting psus (http://www.apc.com/site/support/index.cfm/faq/ ). To avoid this, the ups needs to be sized properly in proportion to the load it is supporting.

Although not necessarily needed, you get a lot more monitoring and management features with the smart-up units and of course they are needed for loads over 2200VA.

As far as I know, APC does not own Cyberpower, however at one time, they did own another UPS company specifically to present with consumer with more buying options. Also it should be noted that although their parent company Schneider Electric owns MGE, all of the MGE single phase product line (1500VA-5000VA) was sold to another company (pretty sure it was Eaton).

Also the return battery capacity value in the screenshot was the minimum battery charge the ups needed before it would turn on after a power outage.

@JGreco
that old smartups you have will last forever as long as your keep putting batteries in it. Those units, specifically from 97-02,03 were over engineered and pretty rock solid. Not saying the ones after that were bad, they just might not last 7+ years.

If you replace the batteries in those SU units and get flashing battery bar graph, you need to recalibrate the battery constant. Tech support used to be able to walk the customer through putting the unit in factory mode and resetting it via CLI, but they stopped doing that a few years ago and would only send out a serial key fob.

 

[jgreco]

Yeah, well, we've had varying experiences with the SmartUPS units. Having had dozens of them in production, I can tell you that some of them had a terrible tendency to do all sorts of things. We had a batch of 1400NET's where the fans were stuck on ("not a problem!" per tech support but we did end up needing to replace those fans periodically), we had units that would overcharge the battery, leading to burning bulging battery packs after a year, we had units that would undercharge the battery, we had units that wouldn't switch over soon enough regardless of the sensitivity setting, and my biggest joy, a frickin' SmartUPS 3000 that appeared to work fine at first glance (testing under constant load). But our environment pairs UPS's with a rack auto transfer switch, and we ended up discovering that this damn unit, which was the secondary, would barf when the RATS decided it was time to transition the load from primary to secondary. The voltage would just sag all to hell for about a second, wiping out the load. I mean fine the power was out and it was supporting no watts and then suddenly it's supporting 8 amps of load but that should be no problem. Bah.

On the other hand, if you kept metrics on your batteries and monitored for problems with a DVM, yeah, you could find the magic units that'd work forever. The unit quoted above is one of them, and it hasn't had a new battery in over 5 years.

With the advent of virtualization I switched us over to the SMT's, which do not seem to burn up their batteries. Been protecting the office rack with a trio of SMX's, I think two 750's and a 1500 with an extra battery, with two RATS, which gives total redundancy. We have had a SMT750 refuse to start due to bad battery after a power outage but I think that, and lack of firmware upgradeability for some of the older ones, are the only issues I can think of.

 

[no_connection]

I have two SmartUPS 1000 get burning hot as one cell in a battery shorted and it just kept dumping power into it. Could be over charge that broke the battery but I didn't check.
The case got so hot that you could barely touch it.
The UPS connected to a Qnap didn't generate any error on the Qnap.
Battery LED indicated and it was beeping, pure luck that I went into the server room and saw it. Even more luck that I could feel the heat.
Could have gone bad or very bad had the shelf not been well ventilated.

The other UPS connected at home with 3rd party software registered that the UPS lost connection to the battery a few times though.
Went home and checked it, problem started the same day for that one so it hadn't gotten as hot.

*edit*
I have four of them, all with bad batteries. Not sure what I should do with them.
If they get burning hot I can't exactly trust then to not burn down the house.

 

[wrath]

Is it still recommend to get a UPS as it seems one company owns all the brands (APC, cyberpower, Eaton) that work with freenas. It seems to me like a single point of failure. It kind of makes all the server grade equipment (ecc ram etc.) and raidz2 useless.

 

[Robert Trevellyan]

It kind of makes all the server grade equipment (ecc ram etc.) and raidz2 useless.

I don't follow your reasoning. If a UPS works properly and protects your system, why does it matter that other brands of UPS are owned by the same parent company, and how does it nullify the benefits of other good practices?

 

[wrath]

I don't follow your reasoning. If a UPS works properly and protects your system, why does it matter that other brands of UPS are owned by the same parent company, and how does it nullify the benefits of other good practices?

I read this thread. I could shell out $150-200 for UPS but it seems that recent hw revisions to same UPS models are low quality and cannot be returned. They have changed behavior when power comes back on before battery is drained, the UPS still shuts down. When people called support, they were told to buy more expensive ups. If they were to buy the same UPS few years back, this problem wouldn't exist.

 

[jgreco]

So your suggested solution, instead of buying a more expensive UPS that has the requisite features, is to say the heck with it and skip the UPS, maybe skip the ECC RAM, and maybe other stuff as well?

 

[danb35]

Why did you necro a thread that had been dead for a year and a half to ask this question? And what does common ownership of the UPS brands have to do with anything?

No UPS is going to do anything to counteract the benefits of ECC, RAIDZ, ZFS, etc. Any UPS will help with power stability. A better UPS will help more, and there may be some which behave less than optimally, but any UPS will be a net gain over no UPS at all.

 

[wrath]

Sorry if not clear, I should have not have brought up ecc ram. If the ups is low quality, with all the redundancy in the world, your data won't be safe. I don't know which UPS to buy because "the manufacturer" keeps cutting corners and masking inferior product as same model and price. I have searched and found like 4 brand names people got working. I don't care about brand names but I have to look because I don't want to buy incompatible hardware.

 

[danb35]

Well, again, virtually any UPS will be better than no UPS. I'm not sure what vendors you have available on the moon, though, so it's hard to recommend anything in particular. I'm using a CyberPower CP1500AVRLCD, which I've had for about a year and has worked well for me. I haven't, though, had any outages long enough to significantly drain the batteries, so I can't say from experience how it behaves in that situation.

 

[Ericloewe]

it seems one company owns all the brands (APC, CyberPower, Eaton)

I'm sorry, but you are either profoundly misinformed or simply trolling. APC is owned by Schenider Electric, Eaton is a big company with nothing to do with Schneider Electric and CyberPower seems to be independent as well.

 

[BigDave]

I'm not sure what vendors you have available on the moon, though, so it's hard to recommend anything in particular.

The question is which moon (in our solar system)

 

[wrath]

I'm sorry, but you are either profoundly misinformed or simply trolling. APC is owned by Schenider Electric, Eaton is a big company with nothing to do with Schneider Electric and CyberPower seems to be independent as well.

You are right I must have misread what DrumNBisco said. Sorry for reopening old thread.

edit: also found compatible hardware list: http://www.networkupstools.org/stable-hcl.html