The Kube Descheduler Operator has a new `DevKubeVirtRelieveAndMigrate` profile.  To leverage this profile the documents include a MachineConfig to set `psi=1` as a kernel argument.

As a resolution of OCPBUGS-37271 we set as a platform default `psi=0`.  These are present in the generated MachineConfigs as a default:
97-master-generated-kubelet
97-worker-generated-kubelet

This causes a conflict where both arguments are present when querying `/proc/cmdline` - See attached cmdline-output.png

When the Descheduler is triggered, it results in an error alert "DeschedulerPSIDisabled" - see attached DeschedulerPSIDisabled.jpg

    As of 4.17

    Consistently

    1. Create a MachineConfig for enabling PSI as per the documentation
---
apiVersion: machineconfiguration.openshift.io/v1
kind: MachineConfig
metadata:
  labels:
    machineconfiguration.openshift.io/role: master
  name: 99-master-kargs-psi
spec:
  kernelArguments:
  - psi=1

    2. Create the example KubeDescheduler CR
---
apiVersion: operator.openshift.io/v1
kind: KubeDescheduler
metadata:
  name: cluster
  namespace: openshift-kube-descheduler-operator
spec:
  managementState: Managed
  deschedulingIntervalSeconds: 30
  mode: "Automatic"
  profiles:
    - DevKubeVirtRelieveAndMigrate
  profileCustomizations:
    devEnableSoftTainter: true
    devDeviationThresholds: AsymmetricLow
    devActualUtilizationProfile: PrometheusCPUCombined

    3. Create a VM or few.

    4. Trigger an action on the descheduler.  This can be done with a CPU Load pod:

---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: cpuload
spec:
  selector:
    matchLabels:
      app: cpuload
  replicas: 1
  template:
    metadata:
      labels:
        app: cpuload
    spec:
      nodeSelector:
        # change to the hostname of the node running the VM
        kubernetes.io/hostname: worker-1
      containers:
        - name: container
          image: 'quay.io/simonkrenger/cpuload:latest'
          resources: # may need to change limits or replicas to trigger load shedding descheduler action
            limits:
              cpu: '8'
              memory: 1Gi
            requests:
              cpu: '8'
              memory: 1Gi
  strategy:
    type: Recreate    

    Descheduler fails, produces error and alert

    VM to move to a more suitable host and rebalancing load

I wonder if this `psi=0` is a too-wide-spread platform default, and the original bug was due to the use of realtime kernel in Telco?

If not, then our documentation for `DevKubeVirtRelieveAndMigrate` profile in Kube Descheduler should include a note about the default?

Thoughts?