Uploaded image for project: 'OpenShift Bugs'
  1. OpenShift Bugs
  2. OCPBUGS-32473

With workload partitioning enabled, setting cpu_manager to static and having reserved cpu causes kubelet fail to restart

XMLWordPrintable

    • Icon: Bug Bug
    • Resolution: Unresolved
    • Icon: Critical Critical
    • None
    • 4.16.0
    • Node / Kubelet
    • Critical
    • No
    • CNF Compute Sprint 252, CNF Compute Sprint 253
    • 2
    • Rejected
    • False
    • Hide

      None

      Show
      None
    • The flow fixed by this bug was never exposed to users. The change of functionality will be covered by OCPBUGS-32472
    • Release Note Not Required
    • In Progress
    • Hide
      2024-04-10: Merged, will be verified once OCPBUGS-28545 merges.
      2024-03-20: Requires attention by maintainers (but ShiftWeek) to merge
      2024-03-11: Currently blocked on CI, the patch is ready
      Show
      2024-04-10: Merged, will be verified once OCPBUGS-28545 merges. 2024-03-20: Requires attention by maintainers (but ShiftWeek) to merge 2024-03-11: Currently blocked on CI, the patch is ready

      This is a clone of issue OCPBUGS-31348. The following is the description of the original issue:
      —
      This is a clone of issue OCPBUGS-29520. The following is the description of the original issue:
      —
      Description of problem:

       On system with workload partitioning enabled setting cpu_manager to static and having reserved cpu causes kubelet fail to restart

      Version-Release number of selected component (if applicable):

      4.16.0-0.ci-2024-02-13-072746    

      How reproducible:

      Everytime

      Steps to Reproduce:

          1. Enable workload partitioning
    2.  Label one of the worker nodes as worker-test
    3.  Create a mcp for worker-test node.
    4.  Create a kubelet config as show below:
apiVersion: machineconfiguration.openshift.io/v1
kind: KubeletConfig
metadata:
  name: kubelet-test
spec:
  kubeletConfig:
    cpuManagerPolicy: static
    reservedSystemCPUs: 0,2,12,14
  machineConfigPoolSelector:
    matchLabels:
      machineconfiguration.openshift.io/role: worker-test

  5. Node goes to NotReady,SchedulingDisabled mode:

      Actual results:

      Node goes to NotReady,SchedulingDisabled mode
[root@cnfdr22 tmp]# oc get nodes
NAME                                             STATUS                        ROLES                  AGE    VERSION
ocp-ctlplane-0.libvirt.lab.eng.tlv2.redhat.com   Ready                         control-plane,master   116m   v1.29.1+2f773e8
ocp-ctlplane-1.libvirt.lab.eng.tlv2.redhat.com   Ready                         control-plane,master   116m   v1.29.1+2f773e8
ocp-ctlplane-2.libvirt.lab.eng.tlv2.redhat.com   Ready                         control-plane,master   116m   v1.29.1+2f773e8
ocp-worker-0.libvirt.lab.eng.tlv2.redhat.com     Ready                         worker                 100m   v1.29.1+2f773e8
ocp-worker-1.libvirt.lab.eng.tlv2.redhat.com     Ready                         worker                 100m   v1.29.1+2f773e8
ocp-worker-2.libvirt.lab.eng.tlv2.redhat.com     NotReady,SchedulingDisabled   worker,worker-test     100m   v1.29.1+2f773e8

      Expected results:

          Node should not go in to NotReady, SchedulingDisabled mode.

      Additional info:

            msivak@redhat.com Martin Sivak
            openshift-crt-jira-prow OpenShift Prow Bot
            Sunil Choudhary Sunil Choudhary
            Votes:
            0 Vote for this issue
            Watchers:
            5 Start watching this issue

              Created:
              Updated: