[isoprophlex]

Few things beats a cheap, powerful Hetzner server, IMO. I host a LOT of stuff on my single €40/month box. 20 core, 64 gb ram, 1 TB ssd, unmetered network.

They have GPU-backed servers too, obviously theyre more expensive, and not ideal for your usecase maybe.

Modal is a good serverless alternative if you want to scale to zero and be able to handle spiky loads.

[api]

Cloud markup over bare metal and wholesale bandwidth is astronomical. Like hundreds to thousands of times in some cases. If you don’t actually need the managed services and other capabilities offered by big cloud, it’s down right stupid to use it.

[tstrimple]

Crazy to me that folks still recommend Hetnzer considering how easily they will shut off your services with zero recourse. They include vague things in their T&C like "2.7. Furthermore, we reserve the right to terminate the contractual relationship without notice for good cause."

There have been plenty of stories just on HN about this. Not a place I'd choose to host anything important.

[lillecarl]

It's always the same story, people get their sever hacked and blame hetzner for not treating them like princesses. They're a low cost provider, keep backups elsewhere (Backblaze) so you can restore to another provider if you have to. Or keep your infrastructure within T&C and they won't cancel you.

[dcminter]

Sounds like that might work for OP while they're using APIs, but if they decide to run models locally can you get GPU equipped servers from Hetzner?

Edit: Not sure if I misread, or if you edited, but I see you addressed this, and yes in that case sounds like this would be a good option if that kind of price range is ok.

[isoprophlex]

Yeah i edited after a minute indeed, thought I was quick enough for anyone to notice haha.

But you really have to do the calculus, as you buy their cheapest servers for at least a month at a time. And with fluctuating loads, a serverless approach might actually be better.

See also: replicate or runpod. But I like modal best; their SDK is pretty comprehensive and has a sane developer experience, replicate is too constrained and runpod instances were unavailable for me half the time...

[akudha]

How smooth is the signup process? I remember reading they require passport copy etc, which makes me uncomfortable

[api]

It’s because they get a lot of fraud. The alternative would be to raise their prices to compensate for theft and the cost of dealing with abuse.

VPS providers like Vultr and Digital Ocean are a little more expensive but a little easier in some cases. They’re still way cheaper than big cloud.

[chrisweekly]

Good experience here w/ Vultr.com (tho v lightly used). It was Derek Sivers' recommendation. https://sive.rs/ti

[krab]

Not in my experience (I have a EU address and payment card).

The sign up was smooth and easy.

[akudha]

I have an American address and card. I suppose I have to provide extensive documentation then?

[krab]

Not sure. Maybe only if the account is flagged somehow? I just registered - there's a simple form that takes a card in the end. And that was just it. I could order a server.

[isoprophlex]

They require extensive proof of identity. I don't mind, I have a EU business so I registered through that... but there's a barrier there for sure.

[thatwasunusual]

How do you handle CI/CD-style automatic deployments with no downtime on dedicated servers? I'm looking for The Perfect (tm) solution myself.

[RadiozRadioz]

Put everything behind a reverse proxy, then set up automated blue/green deployments. It's extremely trivial for OP's scale of application.

[tomwojcik]

Hn moment. Blue green with zero downtime on a VPS is not extremely trivial. Simple for some, at best.

[RadiozRadioz]

Actually, a while back I recorded myself setting this up from scratch (to make a timelapse). I used Nginx reverse proxy, had my apps in Docker containers across my VPSs, the only other tooling I used was Woodpecker CI for the SCM integration and Ansible for my deployment playbooks.

I looked at the raw footage, it was about 45 hours total. So roughly a working week. You can certainly do it faster, I was not rushing and included unnecessary features. There was also plenty of non-CI/CD stuff in that recording.

So yes, I consider this to be trivial. There is so much open source software to lean on these days that does practically everything for you. There are even ready-made k8s distributions (k3s) that will do more in an arguably more standardized way.

People are afraid of it because platforms like AWS/Vercel/etc have coddled them and made them think this stuff is magic. It's just systems, you can do it.

[olyjohn]

Um This doesn't even sound like a use case where 100% uptime is totally necessary. Hn moment indeed.

[cedws]

You could try Basecamp’s Kamal. I can’t vouch for it because I haven’t tried it, but I expect it would be miles better than setting up a gigantic Kubernetes stack and dealing with all that complexity.