In order to gain better results when using Auto Scaling you should have a brief understanding of memory (RAM) usage in Linux-based environments.

Linux does not immediately release RAM and at this point the monitoring system may trigger an action that isn’t necessarily a must. The data is cached in RAM to be ready for a much faster access every next time it is needed. So right after the VPS is started RAM usage will be growing gradually and not freeing up as you might have expected.

What you really need to monitor is swap memory. When all available RAM is filled up and there’s still some data to cache for fast access the system will start using swap.

So in general you need to enable Auto Scaling when swap, not RAM is exhausted because it’s swap indicating the ‘real’ lack of memory.

We recommend not to enable ‘if RAM usage…’ option, but e.g. ‘if Swap usage is above 90% for 10 minutes, add 2 permanent nodes’ instead.