The more interesting question is how long AWS and retail services would continue running and how they would fail. I don't think anyone could say. All people worldwide vanishing in an instant isn't a realistic scenario to plan for—if it happens, then the world has bigger problems and in any case, who would be around to execute the recovery plan?
Still for the sake of the thought experiment...I'm assuming in this scenario the world sees fit to keep supplying the data centers with power and other utilities, but no one has access to the buildings, tries to get in, bomb them etc. In this case, I would expect things to fail gradually, then suddenly and catastrophically, in the period of a few days to a few months. Failure should happen mostly independently by availability zone.
The domain has just been renewed for 5 years. The VPS is paid by credit card, and is set to auto-renew. This should last until the card expires in early 2027. There's money in that bank account so it should be fine.
Then Cloudflare will serve the website for a bit longer, probably a month or two.
After that, only archive.org copies of the website will remain, but at least the pages are self-contained (by design) and will run just fine when served from an archive.
So until about 2027.
There would be issues though. Information would grow outdated, links would break, mailboxes would get filled, etc.
The tax office would probably cause problems if I stop doing bookkeeping though. My tax advisor could delay their wrath, but eventually they'd freeze my account and payments to the web host would stop. I give it a few months, tops.
While I have seen some cloud infrastructures running on life support with minimal supervision for years, AWS tends to shut down, deprecate or migrate backing services every now and then.
ASGs stop autoscaling. RDS forces version upgrades. EBS volumes go poof. Kube APIs vanish. Upstream AMIs, docker images or apt packages get moved to different repo URLs. Sooner or later some auto upgrade or scaling group would go awry and take out a persistent data store or cause services to fail.
There’s also the possibility for an unpatched vulnerability to get exploited.
I give it about 3-6 months, depending on the architecture pattern.
But actual lights? In shared office bldgs it actually depends on the renter's requests to the cleaning service. I've worked many places where the crew would turn the light on/off when entering/leaving (though would be shocked seeing me working in the dark); but I've worked a place or two where they requested all lights stay on all night, because "when mgmt shows up early in the morning they can't find the light switches next to the door" (made up quote but you get it)
The data sources we use do change format from time to time, so I assume one of those would break eventually.
So maybe 2-5 years? :D
with k3s the cluster iself can autoupdate and with the right settings it can update the system too.
PS: running does not mean its secure, just means that it wont die right away :D
It really makes me think that apocalypse / societal collapse scenarios are likely to be more survivable than we think, particularly in Year 1.
Oh, you didn't mean the actual building lights ...