Back to bug 2116980

Who When What Removed Added
Artom Lifshitz 2022-08-09 16:35:08 UTC Target Milestone --- beta
Red Hat One Jira (issues.redhat.com) 2022-08-09 16:35:50 UTC Link ID Red Hat Issue Tracker OSP-18131
Artom Lifshitz 2022-08-09 17:19:32 UTC Component openstack-nova documentation
QA Contact nova-maint rhos-docs
Keywords Triaged
Assignee nova-maint rhos-docs
Artom Lifshitz 2022-08-09 17:19:50 UTC Summary [RHOS-17 beta known issue] vGPU mdev instances are not being cleaned up after guest deletion [17.0 beta known issue] vGPU mdev instances are not being cleaned up after guest deletion
Artom Lifshitz 2022-08-09 17:20:48 UTC Severity unspecified medium
Priority unspecified medium
James Smith 2022-08-09 18:37:20 UTC Doc Text After you delete multiple guests that use vGPU instances, the mdev's remain claimed and are not removed. As a result, the Placement service under-reports the available vGPU resources, which artificially reproduces capacity.
CC astillma, jamsmith
Flags needinfo?(alifshit) needinfo?(astillma)
Doc Type If docs needed, set a value Known Issue
Andy Stillman 2022-08-09 18:48:11 UTC Doc Text After you delete multiple guests that use vGPU instances, the mdev's remain claimed and are not removed. As a result, the Placement service under-reports the available vGPU resources, which artificially reproduces capacity. After you delete multiple guests that use vGPU instances, the mdev's remain claimed and are not removed. As a result, the Placement service under-reports the available vGPU resources, which artificially reduces capacity.
Alex Stupnikov 2022-08-10 07:31:40 UTC CC astupnik
Artom Lifshitz 2022-08-10 18:56:29 UTC Flags needinfo?(alifshit)
Artom Lifshitz 2022-08-10 20:14:31 UTC Keywords ReleaseNotes
James Smith 2022-08-11 02:39:00 UTC Doc Text After you delete multiple guests that use vGPU instances, the mdev's remain claimed and are not removed. As a result, the Placement service under-reports the available vGPU resources, which artificially reduces capacity. When a guest with a vGPU is deleted, migrated off its compute host, or stopped, the vGPU's underlying mdev device is not cleaned up. If this happens to enough guests, all available mdev devices will be consumed, and no further instances with vGPUs can be created on that compute host.
Irina 2022-08-11 08:54:15 UTC CC igallagh
Doc Text When a guest with a vGPU is deleted, migrated off its compute host, or stopped, the vGPU's underlying mdev device is not cleaned up. If this happens to enough guests, all available mdev devices will be consumed, and no further instances with vGPUs can be created on that compute host. If you launch a vGPU instance in RHOSP 17 you cannot delete it, stop it, or move it. When an instance with a vGPU is deleted, migrated off its compute host, or stopped, the vGPU's underlying mdev device is not cleaned up. If this happens to enough instances, all available mdev devices will be consumed, and no further instances with vGPUs can be created on that compute host.
Artom Lifshitz 2022-08-23 15:44:06 UTC Target Milestone beta ga
Summary [17.0 beta known issue] vGPU mdev instances are not being cleaned up after guest deletion [17.0 ga known issue] vGPU mdev instances are not being cleaned up after guest deletion
Andy Stillman 2022-08-24 14:05:03 UTC Flags needinfo?(astillma)
Artom Lifshitz 2022-09-12 15:21:19 UTC Resolution --- CURRENTRELEASE
Status NEW CLOSED
Keywords ReleaseNotes
Last Closed 2022-09-12 15:21:19 UTC

Back to bug 2116980