Description of problem: Version-Release number of selected component (if applicable): 5.14.0-70.el9.x86_64 openvswitch2.16-2.16.0-42.el9fdp.x86_64.rpm dpdk-21.11-1.el9.x86_64.rpm How reproducible: Steps to Reproduce: 1.Run ovs dpdk vhostuser pvp performance test. Actual results: The 4queue 8pmd case got very low performance sometimes. such as got 0.069mpps. But can got 10mmps in manual test. https://beaker.engineering.redhat.com/jobs/6395189 https://beaker-archive.host.prod.eng.bos.redhat.com/beaker-logs/2022/03/63951/6395189/11613965/141138261/bnxt_10.html Expected results: Got the normal performance result(10mpps) every time. Additional info:
For openvswitch2.16-2.16.0-58.el8fdp.x86_64.rpm, 1 queue 2pmd no viommu case got very low performance sometimes, 64byte got 0.89mpps. And 4 queue 8pmd no viommu case got 0. https://beaker.engineering.redhat.com/jobs/6398463 https://beaker-archive.host.prod.eng.bos.redhat.com/beaker-logs/2022/03/63984/6398463/11620060/141192731/bnxt_10.html
For openvswitch2.16-2.16.0-58.el8fdp.x86_64.rpm. run more jobs. some cases got low performance. https://beaker.engineering.redhat.com/jobs/6422502 https://beaker-archive.host.prod.eng.bos.redhat.com/beaker-logs/2022/03/64225/6422502/11658833/141514614/bnxt_10.html https://beaker.engineering.redhat.com/jobs/6421536 https://beaker-archive.host.prod.eng.bos.redhat.com/beaker-logs/2022/03/64215/6421536/11657089/141501748/bnxt_10.html
Please check if this is still the case with OVS 2.17 or 3.1 ? If so, Is this occurring with other NICs on this DUT ? When were these tests last stable and what components have changed since then ? What are the commands used?
I will check if this issue still with ovs2.17 and 3.1. When I run job on rhel9.2, the trex port will down after the connect port bound to vfio-pci. I have open another bug.https://bugzilla.redhat.com/show_bug.cgi?id=2220900