Rebooting a node changes the CPU cores pinning in the pod in SingleNodeOpenShift

Issue

• Rebooting a node changes the CPU cores pinning in the pod in SingleNodeOpenShift.

• The clusters feature is following.

  • The CPU cores defined NUMA0(even number) and NUMA1(odd number).
  • PerfomanceAddonOperator is installed, and numa.topologyPolicy is set best-effort.
  • The pod that CPU cores pinning has a multus addon NIC configured, and the addon NIC is located in NUMA1.

• When a node is reboot, the CPU pinning in the pod changes.

  e.g. When pinning 6 cores to a pod) 
  - PinningCPU before node reboot: 1,3,5,7,9,11 (NUMA1 cores)
  - PinningCPU after node reboot: 0,2,4,6,8,10 (NUMA0 cores)
  

• Shouldn't the NUMA1 core that the NIC is located be allocated even after the node is rebooted? Why even-numbered NUMA0 CPUs are allocated?