Horizontal Scaling for python backend

[rapid lotus]

I have tens of thousands of incoming users, and need to do some horizontal scaling since python does not do well with websockets. How would I go about doing this?

ec4e3dfb-c065-4e54-8d02-7ed44d7eba95

[limber fractal]

Project ID: ec4e3dfb-c065-4e54-8d02-7ed44d7eba95

You might find these helpful:

• how do you compare with render.com?
• Migrate from Heroku | Railway
• Baseline vCPU

⚠️ experimental feature

[urban ferry]

If you're using Railway commercially you should be on the teams plan. You'll get help directly from the team there

[rapid lotus]

Thanks for the heads up!
In the meantime, have you had any experience with this kind of scaling?

[urban ferry]

Not personally no. Railway doesn’t do automatic scaling so you’ll have to build up a load balancer service that balances between multiple backend services. Not sure what the best practice is but it should be fairly simple

First result on Google looks decent: https://testdriven.io/courses/http-load-balancer/concepts/

[smoky vine]

wyz made a scaling system, can't find it rn

cc @mild walrus

what was ur scaling thing for railway

[mild walrus]

I archived it because railway updated metrics API and I don't have enough time to maintain that rn