How can I scale the control plane pods CPU and memory resources

[vaibhavd21]

I have a TCP whose CPU and memory requirements for kube-apiserver, controller-manager, scheduler and konnectivity looks like below:

 ~  kubectl get tcp dec29onecp -n kaapi-lab-default-service -o yaml | \
yq e '.spec.controlPlane.deployment.resources' -
apiServer:
  limits:
    cpu: 500m
    memory: 2Gi
  requests:
    cpu: 50m
    memory: 256Mi
controllerManager:
  limits:
    cpu: 100m
    memory: 250Mi
  requests:
    cpu: 20m
    memory: 64Mi
kine: {}
scheduler:
  limits:
    cpu: 100m
    memory: 128Mi
  requests:
    cpu: 20m
    memory: 64Mi

 ~  kubectl get tcp dec29onecp -n kaapi-lab-default-service -o yaml | \
yq e '.spec.addons.konnectivity.server.resources' -

limits:
  cpu: 200m
  memory: 250Mi
requests:
  cpu: 20m
  memory: 64Mi

The deployment gets created with the resources listed in the TCP's spec and we get 3 replicas of the control-plane pods running with the same resources.
Based on the usage, I want to autoscale the resources for this containers. I thought of VPA first to update the deployment but TCP will revert changes back to whatever in its spec.

[prometherion]

I didn't test this, but wondering if you set no resources at all and manage them externally: essentially, you should be able to change resources directly to the given resources and Kamaji shouldn't try to reconcile values its not responsible for.

In the meantime, it would be great if you could share your autoscaling values, and also the metrics you're considering for it.