Hi fellow hosters!

I do selfhost lots of stuff, starting from the classical '*Arrs all the way to SilberBullet and photos services.

I even have two ISPs at home to manage failover in case one goes down, in fact I do rely on my home services a lot specially when I am not at home.

The main server is a powerful but older laptop to which i have recently replaced the battery because of its age, but my storage is composed of two raid arrays, which are of course external jbods, and with external power supplies.

A few years ago I purchased a cheap UPS, basically this one: EPYC® TETRYS - UPS https://amzn.eu/d/iTYYNsc

Which works just fine and can sustain the two raids for long enough until any small power outage is gone.

The downside is that the battery itself degrades quickly and every one or two years top it needs to be replaced, which is not only a cost but also an inconvenience because i usually find out always the worst possible time (power outage), of course!

How do you tackle the issue in your setups?

I need to mention that I live in the countryside. Power outages are like once or twice per year, so not big deal, just annoying.

  • mark3748
    link
    fedilink
    English
    arrow-up
    3
    ·
    4 months ago

    You’re saying that data centers are replacing batteries constantly…just imagine the labor costs on that (and the down time), not even considering the material cost.

    I’m the tech doing the battery replacements. The big boy UPSes are typically a 3-5 year replacement cycle. Something like this:

    (I just picked the last one on my phone so not a great picture, they’re about the size of a small refrigerator)

    On rack mount and desktop style UPSes 18-36 months isn’t unreasonable. Some of the smaller UPSes, like APC 750s, go through batteries even faster. My personal theory is that they just get and stay too hot.

    There is typically zero downtime while servicing any of them, every critical system has redundant power supply and battery replacements usually don’t interrupt power output anyway. It would take multiple failures to cause any sort of significant downtime, and if it would, we just do them during scheduled downtime.