Move operations need to recalculate NUMA topology
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Compute (nova) |
Fix Released
|
Medium
|
Unassigned |
Bug Description
**Important:** This bug only covers cold migration, evacuation, resize and unshelve operation. Live migration is covered by bug 1289064.
I'm running nova trunk, commit 752954a.
I configured a flavor with two vcpus and extra specs "hw:cpu_
I booted up a number of instances such that there was one instance affined to host cpus 12 and 13 on compute-0, and another instance affined to cpus 12 and 13 on compute-2. (As reported by "virsh vcpupin" and "virsh dumpxml".)
I then triggered a live migration of one instance from compute-0 to compute-2. This resulted in both instances being affined to host cpus 12 and 13 on compute-2.
The "hw:cpu_
Changed in nova: | |
importance: | Undecided → Medium |
status: | New → Confirmed |
Changed in nova: | |
assignee: | nobody → Chris Friesen (cbf123) |
Changed in nova: | |
assignee: | Chris Friesen (cbf123) → Bart Wensley (bartwensley) |
tags: | added: live-migrate |
tags: | added: liberty-rc-potential |
tags: | removed: liberty-rc-potential |
tags: |
added: live-migration removed: live-migrate |
Changed in nova: | |
assignee: | Nikola Đipanov (ndipanov) → Sylvain Bauza (sylvain-bauza) |
Changed in nova: | |
assignee: | Sylvain Bauza (sylvain-bauza) → sahid (sahid-ferdjaoui) |
Changed in nova: | |
assignee: | sahid (sahid-ferdjaoui) → Sylvain Bauza (sylvain-bauza) |
Changed in nova: | |
assignee: | Sylvain Bauza (sylvain-bauza) → sahid (sahid-ferdjaoui) |
Changed in nova: | |
assignee: | sahid (sahid-ferdjaoui) → Stephen Finucane (stephenfinucane) |
Changed in nova: | |
assignee: | Stephen Finucane (stephenfinucane) → Pawel Koniszewski (pawel-koniszewski) |
Changed in nova: | |
assignee: | Pawel Koniszewski (pawel-koniszewski) → sahid (sahid-ferdjaoui) |
Changed in nova: | |
assignee: | sahid (sahid-ferdjaoui) → Pawel Koniszewski (pawel-koniszewski) |
Changed in nova: | |
assignee: | Pawel Koniszewski (pawel-koniszewski) → Andrey Volkov (avolkov) |
Changed in nova: | |
assignee: | Andrey Volkov (avolkov) → nobody |
Changed in nova: | |
assignee: | nobody → Gayathri Devi (gayathrik) |
Changed in nova: | |
assignee: | Gayathri Devi (gayathrik) → nobody |
Changed in nova: | |
assignee: | nobody → Stephen Finucane (stephenfinucane) |
Changed in nova: | |
assignee: | Stephen Finucane (stephenfinucane) → sahid (sahid-ferdjaoui) |
summary: |
- migration/evacuation/rebuild/resize of instance with NUMA topology needs - to recalculate NUMA topology + migration/evacuation/rebuild/resize/shelve of instance with NUMA + topology needs to recalculate NUMA topology |
summary: |
- migration/evacuation/rebuild/resize/shelve of instance with NUMA + migration/evacuation/rebuild/resize/unshelve of instance with NUMA topology needs to recalculate NUMA topology |
summary: |
- migration/evacuation/rebuild/resize/unshelve of instance with NUMA - topology needs to recalculate NUMA topology + migration/evacuation/resize/unshelve of instance with NUMA topology + needs to recalculate NUMA topology |
Changed in nova: | |
assignee: | Nikola Đipanov (ndipanov) → Artom Lifshitz (notartom) |
assignee: | Artom Lifshitz (notartom) → nobody |
summary: |
- migration/evacuation/resize/unshelve of instance with NUMA topology + cold migration/evacuation/resize/unshelve of instance with NUMA topology needs to recalculate NUMA topology |
summary: |
- cold migration/evacuation/resize/unshelve of instance with NUMA topology - needs to recalculate NUMA topology + Move operations need to recalculate NUMA topology |
description: | updated |
On a related note, I just did a resize of an instance from a flavor with dedicated CPUs, 2 cpus and no specified numa information to a flavor with dedicated CPUs, 2 cpus and two NUMA nodes.
As part of the resize this ended up triggering a switch to another compute node. Also, the XML file for the instance being started up did not reflect the NUMA settings for the new flavor.
Lastly, as per above the host cpus were not re-evaluated on the new compute node and so ended up using host cpus that were already in use by another instance with dedicated CPUs.