oh hey, this is the first actual blog post

the calamity

so, as you might have noticed, all my sites were down for most of yesterday (2025 may 28), and that most (at time of writing) are still down

well, that's because for some reason when I went to install updates on the alma linux vps hosting those, it ended up dying, becoming unreachable (no ssh, not responding to pings, and oracle's management panel didn't really give any info about the machine, only if it's 'up'), I still honestly don't know why, as there was no feedback before telling it to restart that something was off, and after trying to reboot then certainly there was none

the good thing is that I had made a backup to my pc just before, not of the whole thing, just the /home of the opc user, which is where all the important things were stored really, so at least there was no loss of important data, maybe I could have done a full backup, but I'll be honest, never really found the backup option in oracle cloud, and I was not going to try transfering about 10 GBs of a system image

after the server became unresponsive after rebooting from it, I of course tried rebooting it a few more times from the oracle panel, which no luck, which then led to me to having to delete the instance and create another, big mistake

you know what's the problem of offering a free vps service? lots of people want it. apparently in the region my account was situated (ashburn) their datacenters were out of available resources for new vms or something, and I then also learned that there's people that write scripts just for spam the "create instance" button to try slip in when there's a chance

the scramble

of course I wasn't going to waste my time for who knows how long pressing a button until a new machine was created, so looking for a new vps provider it is, which is quite a complicated task when your current budget was 0, and you have no income at the moment. was quite an annoying few hours of looking over and over again at dozens of sites and lists of providers, my main criteria being one that overs a package with at least 8GB of ram (most of the time the server was idling with about 7GB any any give time) for at most 7 to 8 eur a month (and was willing to contract for longer terms if it means a better deal, as long as the total to pay didn't exceed 300 brl (46 eur), which it often did with most)

then finally landed on one small and quite unknown one, datalix.eu (not an advertisement for them, just glad they exist), which seemed to have some pretty good pricing somehow (feels a bit too good to be true tho), got one of the packages in their sale for 6 months, which I paid about 31 eur. in a few minutes got a vm running with rocky linux and up for doing things with

the climb

after that, then it was spending an hour or two re setting up the server, creating an admin account (the machine starts with a root account), first things first adding neovim¹, then zsh, git, caddy, etc

a side quest

I wanted to use this re-setup time to do things more properly, and one of the things I've wanted is to have a catch all subdomain routing. previously when you were to go on doesnt-exist.dragon-vi.be you would just see the browser telling you there was an error getting to that page, no fancy error, not even a white page with "404 Not Found" then an <hr> and the name of the webserver

site.com {
    # handle
}

admin.site.com {
    #handle
}

*.site.com {
    # handle
}

but there's an issue with that, caddy, wanting everything to be https, doesn't let you have a wildcard subdomain without either using on_demand tls, or a tls { dns ... } section, the latter being preferred, the thing being that to tho that caddy needs to call an api from your nameserver provider, and it only has integrations with a few providers, and of those none (for what I could find) for hostinger, where I have my domain name and use to setup my dns records

so time to look for a separate dns provider (not another registrar, I still wanted to keep my domains on hostinger), and ended up landing on bunny.net (again, not advertising them), why? second most important, they have pretty generous pricing; most important 🐇 (yes, really, I'm the kind of person who makes my contracting decisions based on having cute mascots)

so after some more time of setting them as my nameserver resolver, then copying the dns records from hostinger to them, rebuilding caddy with the proper plugins, changing the caddyfile to use the tls section, now if you go to a subdomain that doesn't exist (which is quite a few still as I still haven't re setup the other services running in the server) it will show at least something other than a browser error page

the conclusion

at least main site is back up (as you're reading this at all), do have a vps on a company that's not oracle, also now outside of the states, with a slightly more complex setup but it still works

rest will get back up with time whenever I feel like it

make backups of things you care about!