+++ This bug was initially created as a clone of Bug #1883554 +++

+++ This bug was initially created as a clone of Bug #1827283 +++

Description of problem:

All in Title

— Additional comment from Franck Baudin on 2020-04-23 14:23:53 UTC —

Partners requests: https://docs.google.com/document/d/18pzAL9I56zE-JOZokqllVUlCJI1oFAwWJmZq2GDg6XU/edit

AMD CPUs are now shipped o CISCO UCS, Dell, HPE servers, and target NFV market

— Additional comment from RHEL Program Management on 2020-08-07 15:03:31 UTC —

This item has been properly Triaged and planned for the release, and Target Release is now set to match the release flag. For details, see https://mojo.redhat.com/docs/DOC-1195410

— Additional comment from RHEL Program Management on 2020-09-03 10:22:22 UTC —

This item has been properly Triaged and planned for the release, and Target Release is now set to match the release flag. For details, see https://mojo.redhat.com/docs/DOC-1195410

— Additional comment from RHEL Program Management on 2020-09-03 18:16:32 UTC —

This bugzilla has been removed from the release since it does not have an acked release flag. For details, see https://mojo.redhat.com/docs/DOC-1144661#jive_content_id_OSP_Release_Planning.'

— Additional comment from Yariv on 2020-09-07 11:07:57 UTC —

qa_ack+ TP for 16.1.2, not planned yest for 16.2

— Additional comment from RHEL Program Management on 2020-09-07 14:17:07 UTC —

This bugzilla has been removed from the release since it does not have an acked release flag. For details, see https://mojo.redhat.com/docs/DOC-1144661#jive_content_id_OSP_Release_Planning.'

— Additional comment from RHEL Program Management on 2020-09-07 14:51:59 UTC —

This item has been properly Triaged and planned for the release, and Target Release is now set to match the release flag. For details, see https://mojo.redhat.com/docs/DOC-1195410

— Additional comment from RHEL Program Management on 2020-09-08 16:40:15 UTC —

This item has been properly Triaged and planned for the release, and Target Release is now set to match the release flag. For details, see https://mojo.redhat.com/docs/DOC-1195410

— Additional comment from RHEL Program Management on 2020-09-09 16:25:08 UTC —

This item has been properly Triaged and planned for the release, and Target Release is now set to match the release flag. For details, see https://mojo.redhat.com/docs/DOC-1195410

— Additional comment from Sanjay Upadhyay on 2020-09-17 13:47:58 UTC —

There is no QA ack for 16.1.2, on this. Please review the target milestone. Seems like the TM was changed from 16.2 to 16.1 z2, is this some automation?

— Additional comment from Karrar Fida on 2020-09-17 18:07:59 UTC —

This BZ is to track the TP release for 16.1.2. I have created 1880120 to track GA work for release in 16.2.

— Additional comment from Ofer Blaut on 2020-09-23 11:51:28 UTC —

Moving to ON_DEV, since there is nothing to test it or any code change

— Additional comment from Ofer Blaut on 2020-09-24 06:40:04 UTC —

We need to validate that AMD is being documented as tech-preview with additional examples if possible

— Additional comment from RHEL Program Management on 2020-09-29 14:48:29 UTC —

This bugzilla has been removed from the release since it does not have an acked release flag. For details, see https://mojo.redhat.com/docs/DOC-1144661#jive_content_id_OSP_Release_Planning.'

— Additional comment from Erwan Gallen on 2020-09-29 14:55:13 UTC —

NFV FID doc: https://docs.google.com/document/d/1ZnK-frg71nqdan_-imzMx3x1KqYq2xg0svUlEFQSfi4/edit?usp=sharing

— Additional comment from Erwan Gallen on 2020-09-30 12:06:41 UTC —

We are in the early stages of testing, but here is part of the problems seen for RHEL, DPDK, and OpenStack Nova:

• Lack of debugging tools for AMD? (something like pcm.x)
• Ability to control prefetcher (which MSR ?)
• PCIe device NUMA affinity is tied to a specific NUMA node, but this may not represent the actual physical placement, as the PCIe device is connected to the IO die.
• To get the best performances for SR-IOV, we must expose all 32 NUMA nodes and have a hard limit set to 8. OVS-dpdk will be cross socket as a hardcoded parameter limits ovs-dpdk to 8 NUMA nodes. The main DPDK issue is that AMD does not contribute to DPDK.
• Because we do not have the notion of a socket in OpenStack Nova, we do not provide a good NIC affinity with multiple NUMA nodes per socket (option to manage distance NUMA showed by numactl -H)
• Exposed PCIe devices in a multi-NUMA VM are not tied to the proper NUMA nodes.

— Additional comment from RHEL Program Management on 2020-11-19 17:13:20 UTC —

Since this bug report now has devel_ack+, the Devel Conditional NAK state is no longer valid and has been cleared.

— Additional comment from RHEL Program Management on 2020-11-19 19:41:08 UTC —

This item has been properly Triaged and planned for the release, and Target Release is now set to match the release flag. For details, see https://mojo.redhat.com/docs/DOC-1195410