Pin memory usage

I'm running a service that scales memory usage unpredictably and quickly, and it crashes in these cases. I believe strongly that this is due to Railway's autoscaling not provisioning enough memory quick enough (this is obviously hard when you do it all very suddenly). I would be happy to pay for the provisioned resources if it were possible, although I worry this isn't available. Any workarounds?
9 Replies
Percy
Percy10mo ago
Project ID: 95c304a2-fe2b-4ad1-bc3f-5296fd26f36c
Ian Woodfill
Ian Woodfill10mo ago
95c304a2-fe2b-4ad1-bc3f-5296fd26f36c
Brody
Brody10mo ago
how quickly we talking? like 0-3gb instantly?
Ian Woodfill
Ian Woodfill10mo ago
yep
Brody
Brody10mo ago
and what kind of crash logs do you get?
Ian Woodfill
Ian Woodfill10mo ago
No description
Ian Woodfill
Ian Woodfill10mo ago
Is there any documentation around how the autoscaling works at an OS level? would be really helpful for understanding how to work with it is it kubernetes or something under the hood?
Brody
Brody10mo ago
afaik it's vertical scaling as in your app is free to use as much or as little resources as possible, there's no adding additional cpus or memory on the fly I've seen services go from 0mb or 3gb of memory instantaneously without crashing @rayofbytes am I correct in saying this?
Ray
Ray10mo ago
Yep Resources are limited according to plan, so that's your upper ceiling for how much you can use Barring that, you can use anything within that range, and we only bill for what you actually use
running a service that scales memory usage unpredictably and quickly
Can you share more about this?