monitoring

KubeVirtVMIOrdinalPodInterfaceName

Meaning

This alert fires when a running virtual machine instance (VMI) still uses the deprecated ordinal naming scheme (net1, net2, and so on) for its secondary network pod interfaces.

VMs started on KubeVirt versions prior to v1.0 use this ordinal scheme, while newer VMs use a hashed scheme (for example, pod914f438d88d). The condition must hold for 5 minutes before the alert fires.

Impact

This is a warning alert with no impact on operator health. The affected VM continues to run normally for now.

The ordinal scheme is being discontinued: on KubeVirt versions after v1.10, live migration of these VMs breaks. A VM left on the ordinal scheme can therefore block node maintenance and cluster upgrades once the cluster is upgraded past v1.10. The VM remains in this state until it is restarted on KubeVirt v1.0 or later, or live-migrated on KubeVirt v1.10.

As a minor, additional limitation, a VM on the ordinal naming scheme does not support NIC hot-unplug.

Diagnosis

  1. Identify the affected VMI name and namespace from the alert description.

  2. Inspect the pod interface names reported in the VMI status. Secondary interfaces named net followed by digits (for example, net1) use the deprecated ordinal scheme:

    $ kubectl get vmi <name> -n <namespace> \
        -o jsonpath='{range .status.interfaces[*]}{.name}{"="}{.podInterfaceName}{"\n"}{end}'
    

Mitigation

On KubeVirt v1.0 or later, restart the VM. The recreated VMI adopts the hashed naming scheme:

$ virtctl restart <name> -n <namespace>

Note: on KubeVirt v1.10, you can live-migrate the VM once instead, without downtime. Check that the VMI is live-migratable:

$ kubectl get vmi <name> -n <namespace> \
    -o jsonpath='{.status.conditions[?(@.type=="LiveMigratable")].status}{"\n"}'

If it reports True, live-migrate it:

$ virtctl migrate <name> -n <namespace>

These commands act on a VirtualMachine. For a standalone VMI, recreate the VMI by deleting it and re-applying its manifest.

After the VM adopts the hashed scheme, the kubevirt_vmi_vnic_info{pod_interface_name=~"net[0-9]+"} series clears and the alert resolves.

If you cannot resolve the issue, see the following resources: