[Feature] Autoscaling of application

kubevela / kubevela

The Modern Application Platform.

https://kubevela.io

Apache License 2.0

6.42k stars 890 forks source link

[Feature] Autoscaling of application #3867

Open scarlet25151 opened 2 years ago

scarlet25151 commented 2 years ago

Is your feature request related to a problem? Please describe.

for some scenario in games, we would like to bind a game pod and some IaaS resource as a whole application, in some time we need to autoscaling these applications.

Describe the solution you'd like

I've noticed that there is a solution of using KEDA for the autoscaling of some containerized based workload. I would rather to know if there is any native design for the autoscaling of applications.core.oam.dev? like just use application as pod and we can have a layer for the autoscaling of it? Describe alternatives you've considered

Additional context

Thank you for your answer in advance.

wonderflow commented 2 years ago

This is related with https://github.com/kubevela/kubevela/issues/3840 by @fourierr , he's working on that. After that issue finished, I think fourierr will give a solution on autoscaling.

fourierr commented 2 years ago

In my scenario, the autoscaler is working on workload(deployment), just like HPA or KEDA.
From your description, the autoscaler needs to autoscaling the application, it seems that a group of workload will be autoscaling simultaneously.
I think this is a brand-new scene is different from KEDA or k8s HPA

scarlet25151 commented 2 years ago

This is related with https://github.com/kubevela/kubevela/issues/3840 by @fourierr , he's working on that. After that issue finished, I think fourierr will give a solution on autoscaling.

Thank you for providing the information. And I think the situation is a little bit different. Just as @fourierr mentions

From your description, the autoscaler needs to autoscaling the application, it seems that a group of workload will be autoscaling simultaneously.

Yes, here we indeed have some terraform/crossplane cr that is composed together, and there is some topology mapping with the workload so they should be scaled up simultaneously.

because it's comes from our internal demand, I will do some POC first. After that I'm glad if we can have some contribution back to upstream.

wonderflow commented 2 years ago

@scarlet25151 Hello, are there any progress on your PoC?

scarlet25151 commented 2 years ago

@scarlet25151 Hello, are there any progress on your PoC?

Hi @wonderflow, for now we have a dedicate controller named applicationset controller internally to managed the autoscaling of applications. And we are in the refactor so it would be adoption to the latest version, it may need some time for the refactor.