This ticket is tracking the QE verification effort for the solution to the problem described below.

 Problem Description: Clearly explain the issue.

The customer reported that the maximum queue size ( `rxq_desc` ) on physical NICs used by OVS-DPDK is currently limited to 4096 in OVS3.3. On the other hand, from upstream openvswitch information, the limit on maximum queue size are removed.

Detail to be found as follow: 
https://github.com/openvswitch/ovs/commit/60906037033fd87913046b74b9f1962a74c04a7f

 Impact Assessment: Describe the severity and impact (e.g., network down,availability of a workaround, etc.).

Bad performance and packet loss occurred due to queue size.

This issue directly impacts tenant workloads running on the customer's OpenStack environment. 

 Software Versions: Specify the exact versions in use (e.g.,openvswitch3.1-3.1.0-147.el8fdp).

• Openstack 17.1.x

• OVS 3.3

    Issue Type: Indicate whether this is a new issue or a regression (if a regression, state the last known working version).

New issue / feature limitation (not a regression)

 Reproducibility: Confirm if the issue can be reproduced consistently. If not, describe how often it occurs.

The issue is consistently reproducible, repeatable.

 Reproduction Steps: Provide detailed steps or scripts to replicate the issue.

NA

 Expected Behavior: Describe what should happen under normal circumstances.

When the RX queue size is increased or set to unlimited, packet drops (`rx_missed_errors`) should be reduced or eliminated under the same workload.

The actual queue size limitation depends on PMD driver. 

 Observed Behavior: Explain what actually happens.

• `rx_missed_errors` continuously increases, matching the value of `rx_errors`.

• Packet loss is visible to tenant traffic even when PMD threads are not fully utilized.

• Increasing RX ring buffer to 8192 does not change behavior because OVS limits the queue size to 4096.

   Troubleshooting Actions: Outline the steps taken to diagnose or resolve the issue so far.

• Increased RX ring buffer to 8192 becomes 4096

• Cross-checked with upstream commit : "netdev-dpdk: Remove limit on maximum descriptors count", which indicates this limitation has already been lifted by upstream. 

   Logs: If you collected logs please provide them (e.g. sos report, /var/log/openvswitch/* , testpmd console)

NA