Multiple Scale Sets for the same Deployment

[or-he-MA]

I've installed The new ARC-controller and the ARC-Scale-set.

Is there anyway to configure multiple Scale Sets within the same configuration?
For example I have a Scale Set that is configured to org "A", with the label ( Scale Set Name ) set to: "local-runner-a".
Now i have another use case, which I need to run "local-runner-b", where all of the other config is the same. Same GH org, same GH app, everything is the same.

Is there a way to configure within the same deployment multiple scale sets? Or I must create another instance of this resource?
Because as mentioned here:

A single installation of Actions Runner Controller allows you to configure one or more runner scale sets.

It's quite confusing...

Also, is there any possibility to specify tolerations/node selector for the Scale Set? I haven't saw anything similar to this, and I have a use case for arm local-runners, and this is how I make sure it will schedule it on an arm machine.

[johnknutsonhc]

@or-he-MA I had similar questions.

In my organization, we deploy a single scale set controller, and then multiple runner sets within that same cluster. We are currently deploying the helm charts with terraform, which I am not certain is the best route, but that's a different topic.

Below are some terraform snippets that may help illustrate how we are doing it.

resource "helm_release" "arc" {
  name             = "arc"
  repository       = "oci://ghcr.io/actions/actions-runner-controller-charts"
  chart            = "gha-runner-scale-set-controller"
  version          = local.helm_release_version
  namespace        = kubernetes_namespace.arc_systems.metadata[0].name
  ...
  # set values here
  ...
}

# note the `for_each` in the resource block below
resource "helm_release" "arc-runner-set" {

  for_each   = tomap(local.runner_sets)
  name       = each.key
  repository = "oci://ghcr.io/actions/actions-runner-controller-charts"
  chart      = "gha-runner-scale-set"
  version    = local.helm_release_version
  namespace  = kubernetes_namespace.arc_runner_set[each.key].metadata[0].name
  ...
  # set values here
  ...

We do also specify nodeSelectors to limit certain runners to certain nodepools, this is what we set on the scale set controller(s):

  ...
  # limit pods to running on the runners node pool where the `runnerset` label matches the value provided
  set {
    name  = "listenerTemplate.spec.nodeSelector.runnerset"
    value = each.value.runnerset
  ...

[or-he-MA]

Thanks, that might help :)
But another issue i'm encpuntring is that i'm running the scale set in K8s mode, and seems as there isn't Docker installed.
This is a big issue since most of my GH Workflows are building Docker images, and according to the ARC documentation, dind mode dosen't allow to create Docker images.

So, I must use my own image for the self hosted runner? With Docker installed?