This website made me reconsider if 99.9999% SLA and always online is really that important for lots of websites. I used to self-host my own server in my own apartment, but gave up on that after my internet provider had their first major outage that lasted almost a week. Also I didn’t want to worry about the fire hazard, or the server going down when I was traveling etc.
But a solar powered server seems interesting, if the batteries could be made fire safe.
I host my personal site at home with a 15-min battery backup. After that, I accept the downtime. Rarely happens and I haven’t had a single complaint when it does.
Having everything open, on and available 24/7/365 sounds nice until you realize the human toll for things which aren’t that necessary and can be planned around. Down time is a good thing.
doesn’t every website go down when there is no power? ;)
I think the point is to run on solar energy. Being blackout resistant is not an end goal.
Edit: Ohh you’re talking about the post title lol
Only if there is no power everywhere. Some run redundancies in multiple locations.
So… You’re saying that if there’s no power the website goes down.
I used to work on the website for a utility company that insisted on ridiculous uptime. Our major deployments risked taking the system down for a few seconds, though in practice that never happened.
As a result, they made us do deployments starting at 3:00 in the morning since that was the least active time on the site. Fuuuuuuuuck that place.
The uptime looks better than GitHubs…
I know this is a typo, but I’m sad Guthub isn’t a food delivery company.
A disappointment equivalent to Piefed.
gut pull sandwich.com
gut commit -m “ate too much”
gut pushdeleted by creator
GitSub
I have never been a fan of anything above 3 nines of uptime even in a business situation.
Your business will not collapse because of 10 min of downtime a week, and if it does that is not technologies fault.
I don’t even bother to track my home setup.
My home server is 100% up, because I never check it and that’s how it works, right?
My home server is 100% up, because I run the checks from the same machine
I didn’t turn it down, so it’s obviously up.
Schrödingers home server
This reminds me of a news story from the start of Covid.
Tanzanian President, “We have zero bases of Covid!”
Rest of the Earth, “Cool beans! How did you do it‽”
Tanzany Prez, “Easy, we just didn’t test for it.”
Trump: “If you stop the testing, the cases will go down. Drink bleach! UV buttplugs!”
USA: “Let’s try another dangerously old guy for a bit, but if eggs are too expensive after a few years we’ll give this guy another shot.”
Your business will not collapse because of 10 min of downtime a week, and if it does that is not technologies fault.
Hospitals? Air traffic control? Power grid operators? Like these are exceptional businesses, but they’re out there.
I agree, but the number of business that have actual need for extreme uptime is insignificant compared to the number of businesses that don’t but think for some stupid reason they do.
Many airports are closed at night.
Good point, but that’s planned downtime.
My brother is an air traffic controller at one of the world’s busiest international airports and I can assure your their system uptime is no where close to 3 9s.
Those do actually go down. Hospitals have to be evacuated after an earthquake, airports force all aircraft to divert or the grid has the sort of outage that gets it’s own Wikipedia page and needs to be black started.
Those do actually go down.
That’s the point I’m making: when those examples go down unplanned, in particular hospitals and the power grid, in particular the examples you gave, those can be catastrophic events. I.e., if we ignore planned downtime, then these businesses really need five lines of reliability during the times they are expected to be up.
I agree with the overall point that the average business doesn’t need five nines of availability, but I wanted to point out that there are definitely businesses that do.
I found the GitHub employee! 😉
Good news, I don’t work for microslop.
Bad news, I work for someone worse… :(Slopbook?
Na, they are one of our customers.
There are plentiful scenarios where 10 minutes of sequential downtime would be a critical issue. Say you’re a telecoms or internet provider - even a one minute delay to a life-saving message can be enough for someone to die.
Furthermore, lots of things are interconnected now. Your business might be able to handle 10 minutes of downtime, but lets say my business relies on 12 different service providers like you, who each have 10 minutes of downtime during a week using your rationale. That could up being 2hrs of downtime each week.
99.999% is about 5 minutes and 16 seconds of downtime per year.
12 services down perfectly sequentially works out to 76 minutes per year, not 2hours per week.
Not saying there aren’t applications where that can’t be catastrophic, but you’re scale is off for three 9 downtime.
His comment was not based on three nines, the OP is the one that brought up 10 min a week
Yes, there are some that do, I should have said that. But the number of orgs that need extreme uptime is so small compared to business that think they do need it but really don’t.
Oh for sure. Not everyone is cloudflare.
Things like Cloudflare should have that.
I don’t even bother to track my home setup
I host home assistant and if anything goes down I’ll notice pretty fast.
“Hey look everyone! We never heard of grid scale batteries!”
This website is like 10 years old from memory, well before grid-scale batteries were widespread (well they still aren’t in many countries).
Right, and I’m older than the Sony Walkman and that’s why I never heard about “internet”.
Grid scale batteries are only a solution for transients, if that. Actual energy storage needs to be scalable like pumped storage hydro which accounts for 95% of all storage capacity worldwide.
Scaling means you buy 2 instead of one.
That is literally the opposite definition to scalable. Scalable things don’t require twice the resources for twice the capacity.
Buy two instead of one is exactly the worst case scalability you can have.
95% of the world grid storage capacity is pumped storage hydro and batteries will literally never be able to compete because water is the cheapest energy storage medium we could possibly have at grid scale. North America was the world leader on PSH in the 1970-2000 but now China is eating our lunch and our governments no longer have the ability to start real infrastructure projects because of crony capitalism.
I can’t believe we’re watching China become the world leader in renewables while we pretend like batteries are a smart way of storing GWhs of power because that’s the most profitable solution to the private sector.
Enjoy your water.
100% uptime since the last time we turned it on!
works every time half of the time!
Well… Sorta works half the time, but we deleted the bad reviews
Companies claiming 99.9999 uptime is an absurd goal for a SLA, and wouldn’t even survive a small internet outage like a DDoS attack or Akamai outage. I remember in the 90s selling the idea of 5 9’s - 99.999% on my resume until some manager who Interviewed me asked me to state the expected downtime. I stopped claiming that number because it’s impossible without full active redundancy in different locations and not dependent on external providers like cloudflare ever going down.
And please don’t argue maintenance windows “cover it” because that’s not how they work in the real world today.
Downtime per year:
- 99% ~3.65 days (5,256 min)
- 99.9% ~8.77 hours (526 min)
- 99.99% ~52.6 min
- 99.999% ~5.26 min
- 99.9999% ~31.5 seconds
For my home lab I’m happy my stuff is online and content I hardly lose access when there’s a power outage thanks to my ups. But if I tracked I’d be closer to 95%. Good enough but I can’t wait till I just don’t care to run anything.
I worked for a “99.999%” up time facility. No one talks about the exceptions that don’t effect that number. Some are out of our hands like weather or a knob w/ a backhoe.
Others are just made up by upper management to hit the numbers for their bonuses.
Also, any non-trivial system has failure modes that are not binary. Like, if your convertible car has the roof stuck down you can drive it, but you will be sad if it rains. So, is the car “up” or not? Similarly, if your site has writes offline for a database upgrade, but reads work fine, is it “up”?
The concept is just advertising and something that makes sense to senior management, but provides no actual useful metrics.
Over a period of roughly one year (351 days, from 12 December 2018 to 28 November 2019) the server was up for 95.26% of the time. This means that we were offline for 399 hours (which corresponds to 16.64 days).
Not bad at all really, at least for non-mission critical sites. I have solar and I really like it. Solar power on two barns, and a lot of my security cams run on solar. I’m not sure if it’s the formatting of this site, or the grainy pictures, but it oddly reminds me of the Trojan Room coffee pot.
I would be curious to know this uptime relative to any other self-hosted sites.
I remember Reddit and Twitter constantly suffering from uptime issues and they were far more decentralized, with modern infrastructure to match.
Is 95.2% all that far out of the realm of expectation?
95% isn’t great. I run my own servers on solar and I don’t think I’ve had more than a hourof unexpected downtime in the last 3 years, excluding the usual update reboots.
Well, the article did go on to provide better stats once they changed out the panel and battery:
Since then, the uptime has been 100% (March to September 2020), even though we have added a lot of content by making the website multilingual. This is due to the larger (brand new) battery but also to the better orientation of the smaller solar panel. The battery’s capacity will decline over time, so the higher storage capacity will make it last longer. If the uptime remains 100% through the winter 2020/2021, we will probably downgrade to a 20W solar panel.
deleted by creator
The part I quoted was one of their many tries. It seems now, after changing out the solar panel and battery, that they have greatly improved over 95.2%.
Well of course you gotta think, Reddit and Twitter were also probably getting a lot more hits than Low Tech Mag 😅
This was cool 5 years ago, now you can get solar panels and a home battery big enough to power this thing 24x7 without an issue, even if the sun was fully blacked out for a week
edit: a 40kwh battery would last with the power draw:
Power usage 2.62W
like over a year
There’s a lot of articles about supply based usage vs demand based energy production on that website. The idea that it turns off every now and then is sort of part of the authors’ intention of showing “we can live with some outages”.
This article in particular is a good one: https://solar.lowtechmagazine.com/2017/09/how-not-to-run-a-modern-society-on-solar-and-wind-power-alone/
Now they can get better hardware, but should they? They take seriously the low tech part, and part of being low tech is fighting against consumerism culture and keeping with what they have, which is wotking fine
A 40Kwh battery is like 10 grand, I’m not sure why you brought that up. I’d be more likely to invest in a second battery and solar panel. It’s been up for over 23 weeks right now, I’d guess a second solar panel and battery could keeping it up nearly 100% of the time, and it’d cost under $100 for a 200wh battery and 30w panel.
I think you entirelly missed the point then
the chemical battery is often the most expensive and least eco part of the scheme.
There’s a nifty project on Mastodon where someone is running an instance entirely on solar (and I think a battery?). They’ve been tinkering with some of the code to do things like optimise images for reduced weight or CPU cost.
The project is called @solaradmin@solarcene.community and the account for the person running it is on a separate instance, presumably so people can message them about problems: @smallsolar@techhub.social.
congrats guys! we killed it!
This is actually great!
Social media sites should do this. There’s never a good reason to be on [Insert Corporate Slop Social Media Website here] at 3AM.Never a good reason, like different timezones, different working/sleeping hours. Theres no better or worse hour for every single person.
It’s not like those major social media sites don’t have some kind of data center (or rent one) in every “major” (population wise) timezone they operate in. Your other arguments still hold true, but I think even then it would be better for all of us if those websites went down every now and then.
How dare you come here with your sound logic and reason that makes sense
It’s honestly impressive what you can do with under 3W of power these days
At one point in time, I was obsessed with getting a Thinkpad to the lowest power consumption possible.
I think I managed to drop it down to ~3w, with the screen and radio fully disabled.
Uptime for home users is useless. More important is recovery, reboot and resume (the 3’Rs)
You server should reboot automatically, storage should restore itself from unsafe shutdown and your services should resume operations like nothing happened.
This means test your reboot, ensure stuff doesn’t break, and so on.
Failing storage after an hard reboot is the hardest.
For network, i have two ISP connections and a an auto switch script Incase one goes down. But honestly it’s overkill and not needed for 99% of home hosters.
Buying some kind of UPS like a battery powered power strip is also nice but keep in mind that require maintenance to replace batteries once every few years (long enough to forget) and so be useless. The ups shall only last enough to perform a controlled server shutdown using NUTS.
