Uploaded image for project: 'Red Hat Advanced Cluster Management'
  1. Red Hat Advanced Cluster Management
  2. ACM-4919

klusterlet containers are not pinned to mgmt cores - no WLP annotation

XMLWordPrintable

    • Important

      (FYI, originally posted as https://issues.redhat.com/browse/OCPBUGS-11604)

      Description of problem:

      klusterlet container in pod klusterlet-xxxxx is not pinned to reserved management cores.

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

      oc get clusterversion
NAME      VERSION       AVAILABLE   PROGRESSING   SINCE   STATUS
version   4.13.0-rc.2   True        False         5d4h    Cluster version is 4.13.0-rc.2

      How reproducible:

      Always

      Steps to Reproduce:

      1. Deploy SNO via ZTP with workload partitioning enabled
      2. Check mgmt pods affinity

      Actual results:

      pods do not have workload partitioning annotation, and are not pinned to mgmt cores

      The klusterlet pod is QoS Burstable which seems to be change from previous OCP 4.13 builds.

      Expected results:

      All management pods should be pinned to reserved cores

      Pod should be annotated with:

      target.workload.openshift.io/management: '{"effect":"PreferredDuringScheduling"}'

      Additional info:

      Additional info:

      klusterlet-xxx pod annotations
metadata:
  annotations:
    k8s.ovn.org/pod-networks: '{"default":{"ip_addresses":["10.128.0.94/23"],"mac_address":"0a:58:0a:80:00:5e","gateway_ips":["10.128.0.1"],"ip_address":"10.128.0.94/23","gateway_ip":"10.128.0.1"}}'
    k8s.v1.cni.cncf.io/network-status: |-
      [{
          "name": "ovn-kubernetes",
          "interface": "eth0",
          "ips": [
              "10.128.0.94"
          ],
          "mac": "0a:58:0a:80:00:5e",
          "default": true,
          "dns": {}
      }]
    openshift.io/scc: restricted-v2
    seccomp.security.alpha.kubernetes.io/pod: runtime/default
    workload.openshift.io/warning: skip pod CPUs requests modifications because pod
      container has both CPU limit and request

###
Automation test result

SNO management workload partitioning

  should have management pods pinned to reserved cpus

Test result
  END STEP: Checking cpuset for all running containers via crictl inspect on container host - /home/kni/dgonyier/cnf-gotests/test/ran/workloadpartitioning/tests/workload_partitioning.go:85 @ 04/10/23 15:26:51.378 (973ms)
  [FAILED] Expected
      <[]ranwphelper.ContainerInfo | len:4, cap:4>: [

          {
              Name: "klusterlet",
              Cpus: "2-55,58-111",
              Namespace: "open-cluster-management-agent",
              PodName: "klusterlet-7bf7dd5b55-k2wxx",
              Shares: 51,
              Pid: 22356,
          },
      ]
  to be empty

            leyan@redhat.com Le Yang
            rhn-support-dgonyier Dwaine Gonyier
            Votes:
            0 Vote for this issue
            Watchers:
            6 Start watching this issue

              Created:
              Updated:
              Resolved: