This epic tracks all the effort needed to deliver the solution related to the bug described below.

 Problem Description: after migrating from OpenShiftSDN to OVNKubernetes in OpenShift 4.16.38 the customer is seeing for the first time counter increasing for the rx_dropped packets in the br-ex interface, the rate of packet drops is very low and is not identified at the moment any impacted traffic, the rate is like 1 packet dropped per minute the 0,005% of the total received packets. The point is that trying to identify the impacted traffic doing a retis capture of 5 minutes is possible to see more than 900,000 drops so it is very difficult to identify which of those 900K packets are those 5 making the counter to increase. 

Impact Assessment: no specific impact identified, only increase of the dropped packets visible from the statistics of the br-ex interface.

Software Versions: openvswitch3.3-3.3.0-74.el9fdp.x86_64

 Issue Type: new issue after migrating from SDN to OVNKubernetes.

Reproducibility: The drop at the br-ex interface are constantly reproducible at all the customer clusters, while the other drops are always visible in any OpenShift cluster.

Reproduction Steps: -.

Expected Behavior: possibility to identify the packets causing the drop counter to increase from retis.

Observed Behavior: Increase of the dropped packets and difficulties to identify them from the retis capture.

Troubleshooting Actions: -.

Logs: retis capture to be shared in private.