Worst-case scenario pay-as-you-go / pricing disaster mitigation


This is my first serious foray into pay-as-you-go cloud hosting, and I’m wondering what I need to watch out for financially. Ideally I would pre-pay for credits and have an alert that says I’m about to run out, but I’m guessing that’s uncommon in the cloud world for reasons I don’t know enough about.

What I’m worried about is some strange event leading to unexpected use of a resource leading to suddenly huge bills (I’ve heard some horror stories).

If I understand correctly,

  • cpu/memory usage won’t exceed the listed monthly price if I don’t enable autoscaling, so some non-IO loop in my program won’t hurt me
  • volumes have to be manually extended, similarly databases
  • outbound bandwidth usage could at worst set me back $0.12/GB (India), so the way to get a $10.000 bill is if someone downloads my 1M header image from India 85.000.000 times?

Are there other possible costs that could happen based on usage?

And how would one mitigate the (unlikely but possible) outbound data problem?

1 Like

For bandwidth, which is about the only thing that one’d worry about, Fly does (did?) send email alerts when they detect abnormally high usage. At least, I got them back in the day.

Also, Kurt did mention that Fly might refund for “mistakes” within reason: How concerned should I be about a surprise bill? - #9 by kurt

You can mitigate this with a rate-limiter in your code, or by proxying static assets through a CDN like Bunny / Cloudflare.

Other than (on-demand) Machines kept busy by DoS, I think bandwidth pricing is about it.

1 Like

Thanks, it doesn’t seem so scary then :slight_smile:

1 Like