Workers AI Error 429 (3040)
I recently noticed a lot of requests failing to the Workers AI. I looked into it and found out that it is returning a 3040 error. According to the docs this means "No more data centers to forward the request to". I'm using "meta m2m100-1.2b".
My Question is if there is any way i can prevent those errors? For example buying the Workers paid plan? Or is that really an Issue with the Workers AI?
11 Replies
3040 from Workers binding is 3040: Capacity temporarily exceeded, please try again. which probably describes it better
I'm on paid and notice them a fair bit, iirc there's still some manual scaling the team has to do from time to time
I don't think the AI team has said anything specific about them, other then just sane retries are recommended
Ah i see. So this means it's an issue with the workers ai and i can't really do anything about it?
As long as you're staying below the wider limits: https://developers.cloudflare.com/workers-ai/platform/limits/
then nothing you can do about these specifically
Cloudflare Docs
Limits · Cloudflare Workers AI docs
Workers AI is now Generally Available. We've updated our rate limits to reflect this.
Alright thanks for your help!
Translation is 720 requests per minute
Oh well im far away from that
I notice those too, just spiking recently
(I have some automated testing, fixed rate of 4 per min, so wouldn't ever hit those)
Interesting. Thanks for your help!
yea, over more time, they shift as usage shifts (and I assume, stuff scales)
Oh i see
So i guess i will implement some retry logic then, so that it retries a few times before giving an error