Customize app auto-scaling settings

Where can we read/adjust scaling policies (CPU/RAM thresholds)?

We have a feature called “advanced settings” that will be integrated into Qovery V3 to set the CPU thresholds for the auto-scaling. In the meantime, it can be configured via the Qovery API.

Let me come back to you to give you the key to use. cc @a_carrano

Hi @rophilogene sound great, can you share an ETA for V3?

We soft launch the V3 for June, but the final release is for September. Would you be in the loop for the soft launch?

@rophilogene yes please :slight_smile:

Hi @rophilogene, checked the advanced settings of one of our apps and this is the API response:

{
	"settings": {
		"deployment.delay_start_time_sec": 30,
		"build.timeout_max_sec": 1800
	}
}

Where can I edited CPU thresholds for autoscaling now? I was expecting something like 50% CPU or 50% RAM.

Cheers, Florian

Hi @FlorianSuchan ,

I need to see with my team to add the option to set the auto-scaling threshold for the CPU. Basically, the CPU parameter that we need to make configurable is this one targetCPUUtilizationPercentage (default: 60). I need to see with @Pierre_Mavro if auto-scaling on memory is possible with the Horizontal Autoscaler (I think it requires to have a Vertical Autoscaler, which is not for the same purpose at all)

I come back to you in a few hours with more info.

Hi @rophilogene thanks for getting back to us on this, I think memory is not that important right now, was just a question if it is possible.

CPU however, is important. And isn’t there also a time component as in “crossing CPU threshold for more than x seconds”?

In the meantime: What can we do to increase the amount of running instances? Reduce vCPU ressource on the process?

Ok :+1: we’ll see what we can do :slight_smile:

It seems to be possible via the stabilizationWindowSeconds

You have 2 choices:

  1. Increasing the number of nodes on your cluster, then you can increase the number of instances without changing the vCPU size.
  2. Reducing the vCPU size of your instances.

Reducing the vCPU size (not too much) is a good practice to avoid over-allocate resources to your instances.

Note: to know what CPU and RAM size you should allocate to your application, I do recommend using Datadog or any monitoring solution.

Hi @FlorianSuchan,

I’m going to take a look in the week and shortly propose you something to customize what you need.

I’ll contact you directly to ensure all options I’ll add will fit your needs.

Pierre

Thanks guys, awaiting your response :slight_smile:

Hi @FlorianSuchan , Horizontal Pod Autoscaler configuration will be available through app advanced settings next week :slight_smile:

1 Like

Hi @FlorianSuchan , I am super happy to announce that the Horizontal Auto-Scaling settings are now configurable via our app advanced settings.

Let me know if you need anything else :slight_smile: