RBAC Issues with Selenium Grid Helm Chart and Keda Autoscaling: How to Resolve or Bypass RBAC Restrictions?

[nadolualex]

Tried to deploy a selenium-grid Helm chart with autoscaling functionality using Keda, but encountered RBAC issues related to rbac.authorization.k8s.io, getting the following error: resource rbac.authorization.k8s.io:ClusterRole is not permitted in project. I observed errors in several roles, including ClusterRole, ClusterRoleBinding, Role, and RoleBinding. I was wondering if there is any way to enable autoscaling for selenium-grid without RBAC because if I disable it, the Keda pods are refusing to start.

[VietND96]

Set this config key to false.
autoscaling.patchObjectFinalizers.enabled

[nadolualex]

If possible, can you try a dry run with helm template, output all YAML rendered (remove secret info if necessary), and provide that output so that I can understand it further?

Couldn't attach the file, hope it helps

apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  labels:
    app.kubernetes.io/name: keda-operator    
    helm.sh/chart: keda-2.15.1
    app.kubernetes.io/component: operator
    app.kubernetes.io/managed-by: Helm
    app.kubernetes.io/part-of: keda-operator
    app.kubernetes.io/version: 2.15.1
    app.kubernetes.io/instance: release-name
--
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  labels:
    app.kubernetes.io/name: keda-operator-minimal-cluster-role    
    helm.sh/chart: keda-2.15.1
    app.kubernetes.io/component: operator
    app.kubernetes.io/managed-by: Helm
    app.kubernetes.io/part-of: keda-operator
    app.kubernetes.io/version: 2.15.1
    app.kubernetes.io/instance: release-name
--
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  labels:
    app.kubernetes.io/name: keda-operator-external-metrics-reader    
    helm.sh/chart: keda-2.15.1
    app.kubernetes.io/component: operator
    app.kubernetes.io/managed-by: Helm
    app.kubernetes.io/part-of: keda-operator
    app.kubernetes.io/version: 2.15.1
    app.kubernetes.io/instance: release-name
--
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  labels:
    app.kubernetes.io/name: keda-operator-webhook    
    helm.sh/chart: keda-2.15.1
    app.kubernetes.io/component: operator
    app.kubernetes.io/managed-by: Helm
    app.kubernetes.io/part-of: keda-operator
    app.kubernetes.io/version: 2.15.1
    app.kubernetes.io/instance: release-name
--
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  labels:
    app.kubernetes.io/name: keda-operator    
    helm.sh/chart: keda-2.15.1
    app.kubernetes.io/component: operator
    app.kubernetes.io/managed-by: Helm
    app.kubernetes.io/part-of: keda-operator
    app.kubernetes.io/version: 2.15.1
    app.kubernetes.io/instance: release-name
--
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  labels:
    app.kubernetes.io/name: keda-operator-minimal    
    helm.sh/chart: keda-2.15.1
    app.kubernetes.io/component: operator
    app.kubernetes.io/managed-by: Helm
    app.kubernetes.io/part-of: keda-operator
    app.kubernetes.io/version: 2.15.1
    app.kubernetes.io/instance: release-name
--
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  labels:
    app.kubernetes.io/name: keda-operator-system-auth-delegator    
    helm.sh/chart: keda-2.15.1
    app.kubernetes.io/component: operator
    app.kubernetes.io/managed-by: Helm
    app.kubernetes.io/part-of: keda-operator
    app.kubernetes.io/version: 2.15.1
    app.kubernetes.io/instance: release-name
--
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  labels:
    app.kubernetes.io/name: keda-operator-hpa-controller-external-metrics    
    helm.sh/chart: keda-2.15.1
    app.kubernetes.io/component: operator
    app.kubernetes.io/managed-by: Helm
    app.kubernetes.io/part-of: keda-operator
    app.kubernetes.io/version: 2.15.1
    app.kubernetes.io/instance: release-name
--
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  labels:
    app.kubernetes.io/name: keda-operator-webhook    
    helm.sh/chart: keda-2.15.1
    app.kubernetes.io/component: operator
    app.kubernetes.io/managed-by: Helm
    app.kubernetes.io/part-of: keda-operator
    app.kubernetes.io/version: 2.15.1
    app.kubernetes.io/instance: release-name
--
apiVersion: rbac.authorization.k8s.io/v1
kind: Role
metadata:
  labels:
    app.kubernetes.io/name: keda-operator-certs    
    helm.sh/chart: keda-2.15.1
    app.kubernetes.io/component: operator
    app.kubernetes.io/managed-by: Helm
    app.kubernetes.io/part-of: keda-operator
    app.kubernetes.io/version: 2.15.1
    app.kubernetes.io/instance: release-name
--
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
  labels:
    app.kubernetes.io/name: keda-operator-certs    
    helm.sh/chart: keda-2.15.1
    app.kubernetes.io/component: operator
    app.kubernetes.io/managed-by: Helm
    app.kubernetes.io/part-of: keda-operator
    app.kubernetes.io/version: 2.15.1
    app.kubernetes.io/instance: release-name
--
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
  labels:
    app.kubernetes.io/name: keda-operator-auth-reader    
    helm.sh/chart: keda-2.15.1
    app.kubernetes.io/component: operator
    app.kubernetes.io/managed-by: Helm
    app.kubernetes.io/part-of: keda-operator
    app.kubernetes.io/version: 2.15.1
    app.kubernetes.io/instance: release-name

[VietND96]

Okay, I understand your situation. I can confirm that those resources come from the KEDA chart; they don't belong to the selenium-grid resource.
When setting autoscaling.enabled is true. The sub-chart KEDA is deployed together. That chart deployed few resources needed (CRDS, RBAC, so on.) on server side.
In this case, you can ask admin to deploy chart KEDA separately then setting autoscaling.enableWithExistingKEDA in selenium-grid chart

[DeliaPavel]

Hello, @VietND96, but this still means we have to be able to create ClusterRoles, right? Because, if I understand correctly, even if we deploy the KEDA chart separately and then set autoscaling.enableWithExistingKEDA, we still have to have permissions to create ClusterRoles for that KEDA chart.

[VietND96]

I can answer that creating ClusterRoles was unnecessary when chart KEDA deployed.
You can try a dry-run helm template with --set autoscaling.enableWithExistingKEDA=true --set autoscaling.patchObjectFinalizers.enabled=false, only resource ScaledObject or ScaledJob created, which is not RBAC.

[nadolualex]

I can answer that creating ClusterRoles was unnecessary when chart KEDA deployed. You can try a dry-run helm template with --set autoscaling.enableWithExistingKEDA=true --set autoscaling.patchObjectFinalizers.enabled=false, only resource ScaledObject or ScaledJob created, which is not RBAC.

When I do this I still need to install the CRDs for KEDA, which automatically creates ClusterRoles - checked it with helm template command and noticed the following lines when applying CRDs:

[clusterrole.rbac.authorization.k8s.io/keda-external-metrics-reader](http://clusterrole.rbac.authorization.k8s.io/keda-external-metrics-reader) serverside-applied
[clusterrole.rbac.authorization.k8s.io/keda-operator](http://clusterrole.rbac.authorization.k8s.io/keda-operator) serverside-applied
[clusterrolebinding.rbac.authorization.k8s.io/keda-hpa-controller-external-metrics](http://clusterrolebinding.rbac.authorization.k8s.io/keda-hpa-controller-external-metrics) serverside-applied
[clusterrolebinding.rbac.authorization.k8s.io/keda-operator](http://clusterrolebinding.rbac.authorization.k8s.io/keda-operator) serverside-applied
[clusterrolebinding.rbac.authorization.k8s.io/keda-system-auth-delegator](http://clusterrolebinding.rbac.authorization.k8s.io/keda-system-auth-delegator) serverside-applied.

[VietND96]

Yes, ScaledObject or ScaledJob is KEDA's CRDS. And, KEDA needs RBAC to create those resources in the cluster. So, the precondition is KEDA installed to the cluster (via KEDA helm chart or its YAML manifest files). Then install chart selenium-grid with autoscaling.enableWithExistingKEDA=true

[Lucky-Pari]

The error arises because Keda requires RBAC permissions to manage resources like ClusterRole and ClusterRoleBinding. Without these permissions, Keda cannot initialize properly, causing issues with autoscaling.

[Lucky-Pari]

To enable autoscaling without full RBAC, you could use Kubernetes' Horizontal Pod Autoscaler (HPA) as an alternative. HPA can handle scaling based on metrics without requiring Keda's RBAC permissions.

[Lucky-Pari]

If using Keda is essential, you can request limited RBAC permissions from your cluster admin or configure namespace-scoped roles, avoiding global permissions like ClusterRole. Ensure Keda operates within the boundaries of your project’s RBAC policies.