Openstack instance cannot unshelve due to NUMA topology

Solution Verified - Updated -

Environment

  • Red Hat Openstack Plataform 17.1

Issue

After FFU from 16.2 to 17.1, migration and Unshelving are failing with NotImplementedError: Cannot load 'pcpuset' in the base class

Resolution

Open a support case to verify.

Root Cause

Instance was managed by a older version of nova-conductor during or after the upgrade.

Diagnostic Steps

  1. Check details of the unshelve event:
openstack server event show <instance_id> <event_id>
  1. Check flavor configuration of instance, especially cpu_policy:
openstack flavor show <instance flavor>
  1. Check resource used of the node.
(overcloud) stack@undercloud:~$ openstack resource provider usage show <compute_id>
  1. Check available resources of compute node.
lscpu
NUMA node(s):        2
NUMA node0 CPU(s):   0-31,64-95
NUMA node1 CPU(s):   32-63,96-127
  1. Check nova logs in the compute node.
NOVA COMPUTE
2024-03-28 01:27:46.234 2 ERROR nova.compute.manager [instance: 64cf331c-4bb2-4383-af3f-0f9e89244cee] NotImplementedError: Cannot load 'pcpuset' in the base class
NOVA SCHEDULER
2024-03-13 08:36:03.596 29 INFO nova.virt.hardware [req-0bdadb7e-6698-4928-0000-44bf4f27231b 8eb7f06ddf014b808dbe3b5597730000 24b4981a057241b596752a4351900000 - default default] Computed NUMA topology CPU pinning: usable pCPUs: [[12, 76], [20, 84], [71, 7], [78, 14], [79, 15]], vCPUs mapping: [(0, 12), (1, 76), (2, 20), (3, 84)]

This solution is part of Red Hat’s fast-track publication program, providing a huge library of solutions that Red Hat engineers have created while supporting our customers. To give you the knowledge you need the instant it becomes available, these articles may be presented in a raw and unedited form.

Comments