Bug 1760385

Summary: [DOCS] Add to 2.1 release notes: Migration not supported when nodes have different CPU models {BZ 1760028}
Product: Container Native Virtualization (CNV) Reporter: Israel Pinto <ipinto>
Component: DocumentationAssignee: Pan Ousley <pousley>
Status: CLOSED CURRENTRELEASE QA Contact: Irina Gulina <igulina>
Severity: high Docs Contact:
Priority: high    
Version: 2.1.0CC: aspauldi, cnv-qe-bugs, danken, msivak, ncredi, pousley, sgordon, vigoyal
Target Milestone: ---   
Target Release: 2.1.0   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: 2.1 Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2019-11-04 15:07:52 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:

Description Israel Pinto 2019-10-10 12:22:21 UTC
Please add to release notes the limitation in BZ 1760028
Migration not supported when nodes have different CPU models

Comment 1 Martin Sivák 2019-10-15 12:30:23 UTC
The description is horribly confusing. The migration will work or be blocked (or stuck in scheduling) as long as the VM is not using CPU passthrough. That means either the VM.spec.cpu.model has to be set or the default cpu model has to be set in the configmap (like here https://github.com/kubevirt/kubevirt/pull/2028)

Comment 2 Martin Sivák 2019-10-15 12:40:30 UTC
Just to make sure we do not simplify that - the machines can have the same CPU physically, but even a microcode update can in theory change one or two flags and prevent migration of VMs with cpu passthrough (host-model).

Comment 5 Audrey Spaulding 2019-10-21 17:32:25 UTC
Martin, 

Have you had a chance to look at Pan's preview build?

Thanks.

Comment 6 Dan Kenigsberg 2019-10-22 12:33:46 UTC
devel_acking per Audrey's request.

Comment 7 Martin Sivák 2019-10-22 13:25:02 UTC
So, the description is almost right. I would add those notes:

1) the default settings is triggering host CPU passthrough behavior, which can be slightly faster wrt performance, but hinders migration as described
2) the ConfigMap change needs to happen before starting the VMs that are supposed to support live migration
3) an example of cpu model name or a way to figure out the proper name (oc describe node X and look at the cpu-family-NAME labels).

Comment 13 Irina Gulina 2019-10-30 21:49:31 UTC
It has been verified few days ago, PR merged: https://github.com/openshift/openshift-docs/pull/16756

Comment 14 Red Hat Bugzilla 2023-09-14 05:44:20 UTC
The needinfo request[s] on this closed bug have been removed as they have been unresolved for 1000 days