EVC

[dya]

I'm trying to change the cluster's “CPU Model” as per the reference below, but I can't confirm if it actually took effect.
https://community.hpe.com/t5/hpe-morpheus-vm-essentials/evc-mode-in-vme/td-p/7259110
https://community.hpe.com/t5/hpe-morpheus-vm-essentials/mixed-cpu-hvm-cluster/m-p/7258874#M1463

I believe the following method should allow me to verify it, but is my approach correct?

root@vme-iscsi1:~# virsh dumpxml guest-win2025 | grep -i “cpu mod”
<cpu mode=‘host-passthrough’ check=‘none’ migratable=‘on’>
root@vme-iscsi1:~#

[dya]

Changing the “CPU Model” in the cluster does not update the definition files (XML) of existing VMs. It only changes for newly created VMs. How can I apply this change to existing VMs?

[Arnout_Verbeken]

Purpose: When migratable='on' is set, libvirt ensures that the CPU configuration exposed to the guest VM only includes CPU features that are migratable. This means it excludes CPU features that are specific to the host's hardware and might not be available on other hosts in a cluster.

Did you try powering off/on the VM's?  I know you need to do something alike in Vmware.  It's just a thing to test without any fundamental reason behind it.
I'll try to find out more.

[Arnout_Verbeken · Accepted solution]

• Stop the VM
• Move it with "Manage Placement" to another host
• Once moved, the xml should already contain the migratable=on setting
• Power on the VM

[dya]

We have confirmed that the changes have been applied to the existing virtual machine.
After changing the CPU model for the cluster, it is necessary to stop the VM and then move it to a different host in that state.
We had tried powering it off and on again on the same host, but we hadn't tried this approach.

■Before changing the cluster's CPU model

■After changing the cluster CPU model (VM stopped → moved to a different host)

[Arnout_Verbeken]

I think you inverted your screenshots.

But good to know that it's solved!

[dya]

Thank you for confirming.

Regarding the screenshot, I tried again and it is correct as shown above.
It was taken when I changed the cluster setting (CPU MODEL) from “host-passthrough” to “host-model”.