[cyborg][ptg]Cyborg vPTG Summary April 2021
xin-ran.wang at intel.com
Sun Apr 25 09:55:41 UTC 2021
Please ignore my previous message, it is sent to wrong person. Sorry about this.
Thanks for all your participation! We've conducted a successful meeting last week.
Here is the aggregated summary from Cyborg vPTG discussion. Please check it out and feel free to feedback any concerns you might have.
We did a retrospective of Wallaby release, including:
* We supports more operations supported for a VM with accelerator attached.
* We introduced new drivers for Intel x710 NIC and Inspur's NVMe SSD Card.
* We implemented a new configuration file allowing more flexible device configuration.
Here's some major discussion and conclusion of Cyborg vPTG. For more details, please refer to the etherpad.
* More nova operation supporting:
- We prioritized the tasks: 1. suspend/resume. 2. cold migration. 3. live migration.
* vGPU support:
- We reached an internal agreement on whole workflow which can be apply as a generic framework for mdev device.
* API enhancement:
Some of the following items requires a new micro-version.
- Add refresh policy check for all APIs.
- Add device profile update API.
- Add ARQ query by multiple instances API.
- Add disable/enable device API, this one requires a spec first.
- Enhance device profile show API with more information.
* Cleanup issue:
- This issue comes from the case where one compute node shutdown accidently, and the accelerator records
in placement and cyborg DB remains as the orphaned resources. We agreed to implement a mechanism to clean
up the orphaned resources, this one also need a spec.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the openstack-discuss