Container process of burstable pods have their cpu affinity not matching with that  containers cpuset.cpus.  

4.16.0
    

everytime
    

    1. Setup a ocp cluster 
    2. Label one of the node as worker-cnf node.
    3.  Create mcp pool for worker-cnf 
    4. Apply the below Performance profile, where 
apiVersion: performance.openshift.io/v2
kind: PerformanceProfile
metadata:
  name: performance
spec:
  cpu:
    isolated: "0-2"
    reserved: "3"
  workloadHints:
    realTime: true
  net:
    userLevelNetworking: true
  nodeSelector:
    node-role.kubernetes.io/worker-cnf: ""
  numa:
    topologyPolicy: single-numa-node
  realTimeKernel:
    enabled: false
4.  Take any burstable pod (For example ovn-kubenode pod)  and select any container in the pod .  Under the containers cgroup . Check the cgroup.procs to see the container process pid
5. Execute taskset -pc <pid> to get the container process 
6. Check cpuset controller interface file cpuset.cpus 
7. Cpu affinity of container process should match with cpuset.cpus 

Actual results:
Currently container process's affinity matches with reserved cpus and doesn't have access to all the online cpus> 

Example: below is the output of crictl ps to get the container id of ovn-controller container
ce64bd4fbea18       136fb7d44ff690480ae9cc8796c31cfc7c0d9c95c092be11180be478c46b803b                                                         52 minutes ago      Running             ovn-controller                       2                   f8802eae98c2d       ovnkube-node-kz9tk

The container process pid of the above container is : 2932:

sh-5.1# crictl inspect ce64bd4fbea18 | grep pid
    "pid": 2932,
          "pids": {

Get the cgroup path of the container:

sh-5.1# crictl inspect ce64bd4fbea18 | grep cgroup
          "destination": "/sys/fs/cgroup",
          "type": "cgroup",
          "source": "cgroup",
        "cgroupsPath": "kubepods-burstable-pod95dc49f5_b84d_4587_a958_1bc7848fd95b.slice:crio:ce64bd4fbea1873be599b12efa704db236a1bc9626a75123c0d21effff4835ae",

Check the cpuaffinity of the pid 2932
sh-5.1# taskset -pc 2932
pid 2932's current affinity list: 3

Change to container cgroup directory:
cd /sys/fs/cgroup/cpuset/kubepods.slice/kubepods-burstable.slice/kubepods-burstable-pod95dc49f5_b84d_4587_a958_1bc7848fd95b.slice/crio-ce64bd4fbea1873be599b12efa704db236a1bc9626a75123c0d21effff4835ae.scope/
sh-5.1# cat cgroup.procs
2932
sh-5.1# cat cpuset.cpus
0-3

Burstable pod's Container Process affinity should  not have reserved cpus and it's cpu affinity should match with cpuset.cpus