Hacker News Reader

Comment by Waterluvian | Hacker News Reader

It's their loss. I have moved off AWS for this very reason. My pet project cannot involve the risk of possibly costing me thousands or more because I made a mistake or it got super popular over night. I'd rather my site just 503.

View on HN

I know it depends on the site/app, but for a hobbyist, what is the biggest gotcha in a "got super popular over night" situation? If I look at the quotas and overages for low-end plans from the following providers, e.g., it's not obvious to me where the realistic bottlenecks are:

* Firebase Hosting with Firestore

* Cloudflare Workers Sites (using KV)

* Netlify (possibly w/ FaunaDB)

I'd rather my pet project returns 503 after going popular overnight than me footing a huge bill. Especially since my pet projects generally don't generate any revenue anyways. This is the most important feature for me and why I went with GCP.

You know what's going to be cheaper and simpler? Get some vultr VPSes. Maybe one for your web server and one for a PostgresDB, and another for a Redis if you need it.

Done. For 98% of hobbyist projects, a single Vultr $5/month node is probably far more than enough. For 99%, three Vultr $10/month instances (web, DB, cache) is probably enough.

Billing alerts + Lambda. It's not two clicks, but there is plenty of CF templates.

Alerts are just alerts. I don’t and can’t monitor alerts 24/7.

That's why you have the Lambda there to scale it down. You don't need to sit there.

Oh okay, I misunderstood.

The alert automatically calls a Lambda function which turns off all your services.

Which puts a burden on the user because this needs to be tested. Also, it's not 100% safe because the user is still accountable.

Since when was a user being accountable a bad thing?

How do you auto test that this actually works and continues working properly

Set a low limit then scale up.

Or they could just offer a limit ;)

Then you need to be on call 24/7 if someone happens to DDOS you.

That's why you have the Lambda there to scale it down. You don't need to sit there.

Auto-self-DoS is what they were getting at!

Billing alerts are not real time; far from it, actually.

So set a lower limit

The best solution to avoid huge accidental AWS bills due to mismanaging AWS services yourself is to manage your own AWS billing alert service?

Yes, because without them knowing where your infrastructure can be killed/what can be deleted in order to reduce costs without completely destroying your business there's no way for AWS to do this for you.

> without them knowing where your infrastructure can be killed/what

So add an interface that will let you specify that somehow for common scenarios? There must be something better than zero help they can offer. Not everyone needs something that can autoscale to Google levels.

I think the idea is that they do help you. They provide alerts and and APIs that can be used to programmatically control all of your infrastructure. So in a sense having a Lambda listen for billing events and respond in a way appropriate to your particular organization may be pretty close to the best solution.

If your credit card bounces, I'm sure they'd have no problem killing your infrastructure

I'm sure at some point, but at that point you're no longer really their customer and I'm sure they're less worried about not completely destroying your work or livelihood.

I've actually had my payments on my personal account bounce once or twice and no, they did not.

there is plenty of CF templates

This by definition is deploying something you don't fully understand. If there's a problem in any of those templates you won't know. You won't really know if they even do what they say they do.

Using one to do something as important as this would be crazy.

I don't think he's saying that you should blindly deploy this stuff. But you don't have to create a solution from scratch. There are existing templates out there that you can leverage to build your own solution

Reply on news.ycombinator.com