Uploaded image for project: 'OpenShift Bugs'
  1. OpenShift Bugs
  2. OCPBUGS-2963

noobaa-core not starting when cpu Manager is configured on cluster

XMLWordPrintable

    • None
    • OCPNODE Sprint 228 (Blue), OCPNODE Sprint 229 (Blue), OCPNODE Sprint 230 (Blue), OCPNODE Sprint 231 (Blue)
    • 4
    • False
    • Hide

      None

      Show
      None

      Description of problem:

      The cluster is configured with cpuManager set to static , and installation of ODF is not succeeding due to noobaa-core is not starting up
      

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

      4.11, may exist on other OCP versions as well. 
      

      How reproducible:

      always
      

      Steps to Reproduce:

      1. Configure Custer with CPUManager set to static:
      
      
      2. Install Local Storage Operator and ODF/OCS operator
      plus configure the volumeSets and volume Discovery and the different components including StorageCluster 
      
      

      Actual results:

      The noobaa-core StatefulSet POD is not starting
      and fails with error 
      Warning   SMTAlignmentError        pod/noobaa-core-0                                                                     SMT Alignment Error: requested 1 cpus not multiple cpus per core = 2

      Expected results:

      noobaa-core-0 pod must start with no error
      
      

      Additional info:

      cat /etc/kubernetes/kubelet.conf | grep -e cpuManager -e topologyManager -e reservedSystemCPUs
        "cpuManagerPolicy": "static",
        "cpuManagerPolicyOptions": {
        "cpuManagerReconcilePeriod": "5s",
        "topologyManagerPolicy": "single-numa-node",
        "reservedSystemCPUs": "0-5,56-61",   
      
      
      --- The applied Performance profile to the nodes with ODF Storage
      apiVersion: performance.openshift.io/v2
      kind: PerformanceProfile
      metadata:
        name: worker-std-r750-1
      spec:
        additionalKernelArgs:
          - intel_idle.max_cstate=0
          - processor.max_cstate=1
          - idle=poll
          - irqaffinity=0-5,56-61
          - rcu_nocb_poll
          - nmi_watchdog=0
          - audit=0
        cpu:
          isolated: 6-55,62-111
          reserved: 0-5,56-61
          balanceIsolated: true
        globallyDisableIrqLoadBalancing: true
        hugepages:
          defaultHugepagesSize: 1G
          pages:
            - count: 32
              node: 1
              size: 1G
        realTimeKernel:
          enabled: false
        numa:
          topologyPolicy: "single-numa-node"
        net:
          userLevelNetworking: true
        workloadHints:
          highPowerConsumption: true
        nodeSelector:
          node-role.kubernetes.io/worker-std-r750-1: ""
        machineConfigPoolSelector:
          machineconfiguration.openshift.io/role: worker-std-r750-1
      
      

              jmencak Jiri Mencak
              lhalleb@redhat.com Lazhar Halleb
              Votes:
              0 Vote for this issue
              Watchers:
              6 Start watching this issue

                Created:
                Updated:
                Resolved: