Customer is running an OCP 4.16.54 cluster where ovn-controller is consistently consuming 100% CPU on all worker and control-plane nodes.

We performed initial data collection and observed the following:

• Network Policies: 1407
• ACLs in OVN NB DB: 7234

• OVS flows (br-int): ~130,774

• Several ACLs contain a very large number of associated address sets

Given the scale of network policies and resulting ACLs/flows, we suspect that OVN rule processing and/or flow programming might be contributing to the high CPU utilization. We need engineering assistance to confirm what's causing this high CPU usage in terms of OVN.

Data to be Shared:

• OVN namespace inspect
• sosreport from one affected node
• OVN NB/SB DB dumps