Bug 1976671 - [RFE]Official support for Openshift on Openstack with GPU.
Summary: [RFE]Official support for Openshift on Openstack with GPU.
Keywords:
Status: NEW
Alias: None
Product: Red Hat OpenStack
Classification: Red Hat
Component: openstack-nova
Version: 16.1 (Train)
Hardware: x86_64
OS: Linux
unspecified
low
Target Milestone: ---
: ---
Assignee: OSP DFG:Compute
QA Contact: OSP DFG:Compute
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2021-06-28 01:40 UTC by youngcheol
Modified: 2023-03-21 19:45 UTC (History)
7 users (show)

Fixed In Version:
Doc Type: If docs needed, set a value
Doc Text:
Clone Of:
Environment:
Last Closed:
Target Upstream Version:
Embargoed:


Attachments (Terms of Use)


Links
System ID Private Priority Status Summary Last Updated
Red Hat Issue Tracker OSP-5485 0 None None None 2022-03-12 08:45:58 UTC

Description youngcheol 2021-06-28 01:40:06 UTC
Description of problem:

CU wants to ues GPU sharing in Openshift on Openstack instance with GPU.

They wants to got official support with guide about it.

I'll put my finding additional comment.

Version-Release number of selected component (if applicable):
RHOCP 4.x
RHOSP16.1

Comment 2 Artom Lifshitz 2021-06-30 15:02:14 UTC
Hello,

I'd like to get some clarifications on the request.

On the OpenStack side of things, we have documentation around configuring PCI passthrough [1], which can be used to pass through any PCI device to an instance, including GPUs.

We also have a chapter on configuring virtual GPUs [2] - the use case is similar to [1], with the main difference that [2] is only for GPUs, and the instance receives a virtual GPU that's a "slice" of a single physical GPU, whereas in [1] the instance receives the entire PCI device.

Give these two pieces of documentation, what's missing to enable your use case? Do you need OpenShift-specific documentation on how to consume a (virtual) GPU and make it available to containers?

[1] https://access.redhat.com/documentation/en-us/red_hat_openstack_platform/16.1/html-single/configuring_the_compute_service_for_instance_creation/index#configuring-pci-passthrough
[2] https://access.redhat.com/documentation/en-us/red_hat_openstack_platform/16.1/html-single/configuring_the_compute_service_for_instance_creation/index#configuring-virtual-gpus-for-instances

Comment 3 youngcheol 2021-07-07 09:43:07 UTC
Hi Artom,

Sorry and Thank you for the update.

CU is wondering that that architecture is full supported by Red Hat. I was confused about it in our internal DOCs.

Could I tell CU it is full supported?


and If it possible they want to some official DOC and architecture info about it. (if it possible)


THank you,
Regards,
YoungCheol.

Comment 4 Artom Lifshitz 2021-07-07 15:36:52 UTC
Hi again,

I'm still not fully clear on the use case. Are you looking for official documentation on running the NVIDIA GPU Operator in an OpenShift GPU worker on top of an OpenStack VM with a passed-through GPU? In other words, an official version of what Kevin describes in his blog post that you linked [3]?

[3] https://kdjlab.com/openshift-4-6-on-openstack-with-gpu-operator-v1-3/

Comment 7 youngcheol 2021-08-02 01:54:15 UTC
Hi Artom,

I'm still not fully clear on the use case. Are you looking for official documentation on running the NVIDIA GPU Operator in an OpenShift GPU worker on top of an OpenStack VM with a passed-through GPU? In other words, an official version of what Kevin describes in his blog post that you linked [3]?

[3] https://kdjlab.com/openshift-4-6-on-openstack-with-gpu-operator-v1-3/

=> yes, We need official documentation about it, and CU asked officially supported installation guide and supportability about it.

Thank you,
Regards,
YoungCheol.


Note You need to log in before you can comment on or make changes to this bug.