I recently came across a “bullet-dodged” post showing deformed UPS batteries and mentions of hydrogen sulphide leaks.
At first, this surprised me, because I didn’t remember there being any warnings with the last UPS I bought in 2016. Checking the manual, it says “3 - 5 years depending on the number of discharge cycles and environmental temperature”. In my case, power pretty rarely goes out, and the UPS is in a basement so it’s relatively cool. It automatically does a charge cycle every 6 months or so. I just took a look at the battery, and there’s no bulging or signs of issues with it.
Yet, this page on APC.com says that you should replace a battery after at most 5 years. Honestly, that’s fine (replacements look to cost about $30 Canadian), but what really surprises is me is that if the safety aspect is so important, why didn’t it say so in the box? This page looks to have first been published in 2012, so I would have thought any warnings would be in a 2016 UPS manual.
Perhaps this means for a home office style UPS it’s OK to wait until the replace battery indicator is flashing?
Best practices these days is not to use a lead acid based UPS at all. Anker makes a line of very affordable lithium based power banks that also feature UPS switching capability. For about $200, you can get a 300W powerbank/UPS with 10ms switching, nice pretty sine wave output, and about 3-4x the runtime of a typical 1500VA lead acid UPS.
I just discovered these a month or so ago, and I’m in the process of replacing every lead acid UPS I’m responsible for with these, as they fail (and in some cases sooner).
Another nice feature: if the Anker Solix powerbank/UPS goes dead flat in an extended outage, it won’t offer power to the connected devices again until it charges to 2%, which takes it a couple of minutes of uninterrupted charging. This is nice, because it makes CERTAIN that even devices that automatically power on the second they see power won’t see power until the power is actually present and stable–so if the linemen repairing your crap do what the linemen usually do, and have your house hot while they work out the last few issues (delivering brownouts, stuttering, and other nastiness down the line) your precious machines won’t see any of it, because if the Solix doesn’t have enough power to protect them from it, the Solix won’t deliver it to them in the first place.
I believe this is configurable with the app–which I have not yet really plumbed the depths of, choosing instead to focus on stuff like watching the handoff happen in real time over an oscilloscope–but that’s entirely possible.
For me personally, that’s not a deal breaker. I’m already used to, and generally prefer, manually restarting systems after an outage extended enough to require a shutdown. It would be nice to have the auto on thing at least available, especially since it doesn’t energize the outputs until it has about 2% recharged, but I can live without it.
I similarly haven’t messed with the notification stuff. It can certainly be done one way or another, since the app has notifications available, but this is again something I’m personally not too concerned with, since I expect and demand all my systems to be crash safe anyway, on the logical side.
To me, the UPS is about maintaining uptime during outages and protecting the physical boxes from damage. I don’t really expect it to keep a crash from ever happening in the first place.
Anker Solix lineup; the C300 is the smallest, at slightly more expensive than a typical 1500VA lead acid UPS but three to four times the runtime (and many, MANY years and discharge cycles greater endurance).
That figure is both from me finding the AH rating on the battery pack in an APC 1500, and comparing it to that of the Anker and to me doing three run flat tests on the Anker using a desktop machine running a moderately chewy workload (mostly, a never ending rate limited fio test) until crash.
You’ll note that I didn’t do a run flat test on an actual 1500VA UPS. That’s because none of mine are brand new, and none of them can match their specified runtime anymore.
So, I’m comparing the theoretical maximum runtime on the lead acid jobs to the REAL WORLD run time on the C300.
I’m wondering how many more lithium battery fires it’ll take to turn public opinion against this technology. Apparently LFP batteries are much better than traditional lithium ion in this regard (though with inferior energy density) but they’re still a mountain compared to a lead-acid molehill.
For those of us already having a Lead Acid based UPS - is it worth replacing the batteries with Lithium-Ion or -Phospate based battery packs when it is time to replace the batteries?
I do not know if it is possible to change the firmware to support the changed battery-type but some packs are intelligent to “translate” the battery management meant for lead-acid batteries to Lithium based.
Maybe if you’re talking about very expensive rack mount UPS gear and such a replacement pack becomes available. For typical, run of the mill “desktop” UPS models selling for <$300 US, I don’t think it’s worth it.
@mercenary_sysadmin Are you just using these power banks for “single desktop” type setups or for more traditional server workloads? Wondering if this might be an option for small branch office locations that only have one or two servers and some POE switches instead of a traditional large rackmount UPS, especially if move up to the C800 or C1000 models? It would certainly be cheaper than a brand name rackmount unit with LiPO4 batteries.
I’m using them for server workloads as well as desktop workloads. I’ve seen nothing to convince me that lead-acid is better for a server workload in any way that matters to me.
Since runtime is my bottleneck WAY before power-handling is, that also means I can get 3-4x the runtime for the same amount of money I’d spend on cheapish 1500VA UPSes. So at this point, basically I’m just making sure I have enough power-handling for the workload, and calling it a day.
The smallest Anker model is the Solix C300 with 300W power handling, but there is a C800 with 800W handling (and even more runtime) that I’m using for somewhat larger workloads; I’ve got a pair of rack-mount servers with a pair of 24" monitors, couple switches, etc in my downstairs office that I’m running from a C800.
Basically, just make sure you know what your power draw is for your workload, and plan accordingly. With lead-acid UPSes, it was VERY unlikely that power handling was going to be your bottleneck; runtime probably would be. With these, power handling may be a bottleneck, so you need to pay a bit more attention to it… but 300W is a hell of a lot for a single machine to hit, unless it’s a gaming machine with a monstrous GPU (or an AI host with an even more monstrous GPU).
The only other potential fly in the ointment is the number of 120VAC outlets per unit. The C300 only has two 120VAC outlets on it, which means in some circumstances I need to pair it with an el cheapo power strip to get more outlets.
In some ways, though, I actually kind of like that “problem.” One of the issues I frequently have with traditional 1500VA+ UPSes is that they have too many outlets, which then ends up with people plugging in vacuum cleaners, coffeemakers, personal space heaters, you name it into the other outlets, which in turn causes serious issues.
When you only have enough outlets to plug in the stuff that needs it, and no more… those issues disappear!
My immediate thought was it had to be a scam, but there sure are a lot of legit-looking reviews on Amazon.
Backing out to looking for non-Amazon reviews using a search engine (Kagi), I’m a little disturbed at just how many astroturfed reviews are out there. It looks like every shitty pay-for-pay blog that takes money to give excellent reviews has taken their money and given them the glowing review they bought. That doesn’t mean it can’t be a good device, but it’s not the best sign IMO.
I might still buy one and evaluate it personally.
edit: pulled the trigger and bought one in white; $171 after tax, including $20 discount offered on Amazon right now. I’ll update y’all next week on whether it’s an obvious scam or not.
From looking around this is probably a very basic generic bulk run from a Chinese factory which will be rebranded by multiple vendors on AliExpress like the battery bank units. Not that means it is necessarily a bad unit, but longevity, warranty and service can always end up being an issue.
crash-safe filesystem (ext4, NTFS, UFS2, OpenZFS all count)
crash-safe database engines (PostgreSQL, InnoDB, MSSQL, not MyISAM or most “serverless” flat-file DB types)
crash-safe applications (meaning they need to wrap DB operations in ACID-compliant transactions)
The last one is the only one that’s difficult to be certain of. In practice, as long as the first two steps are taken care of, the last one is usually going to be a problem only infrequently even if it’s an actual problem.
But it’s pretty hard to be sure ahead of time, because most vendors don’t really go out of their way to talk about whether they’ve actually built their apps crash-safe or not, or what steps they took to do it, if they did.
