Applying a performance profile on an ARM cluster, results with the tuned profile to turn degraded.

    4.18.0-0.nightly-arm64-2025-02-08-033503

Apply a performance profile     

    1. Label a worker node with a custom label (e.g worker-cnf)
    2. Create an MCP referencing that label
    3. Apply a performance profile:

apiVersion: performance.openshift.io/v2
kind: PerformanceProfile
metadata:
  name: performance
spec:
  cpu:
    isolated: 1-3
    reserved: "0"
  hugepages:
    defaultHugepagesSize: 2M
    pages:
    - count: 2
      size: 2M
  machineConfigPoolSelector:
    machineconfiguration.openshift.io/role: worker-cnf
  nodeSelector:
    node-role.kubernetes.io/worker-cnf: ''
  numa:
    topologyPolicy: single-numa-node
  workloadHints:
    highPowerConsumption: true
    perPodPowerManagement: false
    realTime: true

oc get profile
NAME                          TUNED                                    APPLIED   DEGRADED   MESSAGE                  AGE
ip-10-0-1-225.ec2.internal    openshift-control-plane                  True      False      TuneD profile applied.   3d1h
ip-10-0-37-254.ec2.internal   openshift-control-plane                  True      False      TuneD profile applied.   3d1h
ip-10-0-51-111.ec2.internal   openshift-node-performance-performance   True      True       TuneD profile applied.   3d1h
ip-10-0-66-39.ec2.internal    openshift-control-plane                  True      False      TuneD profile applied.   3d1h
ip-10-0-74-241.ec2.internal   openshift-node                           True      False      TuneD profile applied.   3d
ip-10-0-9-208.ec2.internal    openshift-node                           True      False      TuneD profile applied.   3d1h


Taking a look in the tuned pod I found:

2025-02-12 12:36:24,027 INFO     tuned.plugins.plugin_bootloader: cannot find grub.cfg to patch
2025-02-12 12:36:24,028 INFO     tuned.plugins.plugin_systemd: setting 'CPUAffinity' to '0' in the '/etc/systemd/system.conf'
2025-02-12 12:36:25,086 INFO     tuned.plugins.plugin_script: calling script '/usr/lib/tuned/cpu-partitioning/script.sh' with arguments '['start']'
2025-02-12 12:36:25,116 ERROR    tuned.plugins.plugin_script: script '/usr/lib/tuned/cpu-partitioning/script.sh' error output: 'modinfo: ERROR: Module kvm_intel not found.'
2025-02-12 12:36:25,116 INFO     tuned.daemon.daemon: static tuning from profile 'openshift-node-performance-performance' applied
I0212 12:36:25.116677    2683 controller.go:702] tunedRecommendFileRead(): read "openshift-node-performance-performance" from "/etc/tuned/recommend.d/50-openshift.conf"

    The ARM cluster that was tested here was hosted on AWS.
The worker-cnf node is a VM, and we need to invesigate if this is a leading cause to this failure.