In a 120 node cluster envrironment I am seeing some ebpf agent pods are using very high CPU (as much as 14 vCPU), the pod with most CPU usage happens to be worker node. Couple of Observations from below:

1. some pods use steadily very high CPU 
2. The traffic spike for certain pods at 15:00 hours which  is due new traffic that was generated at the time, average flow rate spiked from 20K/sec to 60K/second at the same time while CPU usage climbed ~14