One alert is never enough. Treating alerting as a system, not a single email.
GCP pricing calculator in the deployment checklist, not an afterthought. You almost have to be obsessive about checking.
Hard caps: GCP doesn't have them. Budgets only send alerts, they don't stop charges. You can wire automation to disable billing but that's a hack, not a feature.
Cloud providers benefit financially from complexity and opaque defaults. I own my config mistakes, but the suggestions around capacity planning and realistic cost modeling were genuinely helpful.
Sorry to hear this happened to you. Everyone's piling on, like they've never made mistakes. With no "hard" limits, it's almost like google wants this to happen. Have you tried contacting your credit card company? They may be able to chargeback google. Maybe you can make a case about idle capacity?
I have enabled some google services to play with (maps, calendars), that required me to put my credit card down. I'm thinking I should get a prepaid credit card so if things get out of hand, they'll shut me down rather than keep charging me until my max credit limit.
Photo culling app. Photographers upload 1000+ images, ML clusters similar shots so clients can pick favorites faster. Needs serious parallel compute (but infrequently). Cloud SQL is the baseline (~$40-50/month, doesn't scale to zero).
My fuckup was setting --cpu 4 --memory 16Gi during initial migration and only optimizing instance counts, not resource sizing.
Didn't help that GCP has no hard spending cap. Budgets only send alerts (in my case 1 easy to miss alert saying you spent more than $50 this month), they don't stop charges... Looking at my credit card statement there are multiple $1-2k charges.
The incentives aren't exactly aligned toward helping you spend less.
Appreciate the directness, and fair point. That’s exactly what I find confusing about their setup and why I’m here trying to learn.
A little more context: I’ve been on GCP for 4 years, App Engine for the majority of it. Expensive but stable. I’ve used Gemini in the past to reduce costs successfully, so this wasn’t my first attempt at optimizing.
I take ownership of the outcome, but the config behavior still doesn’t match my mental model and Google support hasn’t been able to clarify how to properly scope this either, which is why I turned here.
> the config behavior still doesn’t match my mental model
Could you -learn- how to self-host a version of your app to expand your mental model in doing so? You outsourced the thinking part to an LLM - a bag of words - and are surprised the outcome didn't just work?
> Google support hasn’t been able to clarify how to properly scope this either
More outsourcing of thinking, no? Is Plan A really asking the vendor selling you compute how to use less compute and make them less money, instead of figuring out how to use just enough of it yourself?
If you're taking ownership, who could have effected the outcome the most here? Maybe the person who keeps outsourcing thinking to LLMs, support requests, and forums? I'd argue ownership would look more like figuring out how to handle Top Cost #1 yourself and reduce burn rate, starting by doing less outsourcing.
A few things that stood out to me:
One alert is never enough. Treating alerting as a system, not a single email.
GCP pricing calculator in the deployment checklist, not an afterthought. You almost have to be obsessive about checking.
Hard caps: GCP doesn't have them. Budgets only send alerts, they don't stop charges. You can wire automation to disable billing but that's a hack, not a feature.
Cloud providers benefit financially from complexity and opaque defaults. I own my config mistakes, but the suggestions around capacity planning and realistic cost modeling were genuinely helpful.