We are working on scaling up our node js service supported by render and had a few issues when doing some load testing so wanted to ask a few questions. We currently have the Starter plan ($7/mo, 512 MB, 0.5 CPU).
Are there hidden limits (aka x requests incoming/out going per second) in our plan?
If there are proxying limits internally, what would happen, x requests would fail without rejection?
How does render’s horizontal scaling work, does it load a mirrored instance of our server or a brand new one?
In the metrics dashboard, how precise are they in relation to CPU and RAM usage in real time and what does it mean by 0.05 CPU (see below)?
There are no “hidden limits” but you are limited by what your RAM and CPU can handle
If you do reach the limit of what your application can handle we will OOM kill your application and restart it. During that time your app will be unaccessible.
Scaling works by creating multiple instances of your application. There is a load balancer in front of service instances to evenly distribute network request traffic. You can also take advantage of auto-scaling to automatically scale your service up and down based on memory or cpu limits.
Each point on the metrics dashboard for memory and cpu is an average of the instance taken over three minutes (we are working on making this more precise). 0.05 is the fraction of a CPU you are using so 1/20th of a CPU.
