Description of problem:

Dual-stack vSphere IPI installs (e2e-vsphere-ovn-dualstack-primaryv6) fail because the generated Cluster API VSphereMachine network devices have IPv6 DHCP disabled dhcp6: false while the MachineConfig expects ip=dhcp,dhcp6, leading to control-plane machines never obtaining IPv6 addresses and remaining in WaitingForIPAllocation.

On vSphere, the VSphereMachine specs under clusterapi_output show:

spec.virtualMachineCloneSpec.network.devices[0].dhcp4: true
spec.virtualMachineCloneSpec.network.devices[0].dhcp6: false

At the same time, the rendered dual-stack MachineConfigs contain kernel arguments:

ip=dhcp,dhcp6

As a result, control-plane Machine and VSphereMachine conditions repeatedly report reason=WaitingForIPAllocation and the installer times out waiting for control-plane provisioning.

Version-Release number of selected component (if applicable):

4.20.0-0.ci-2025-08-28-164352-test-ci-op-kwrgw8wh-latest
    

How reproducible:

Always

Steps to Reproduce:

1. Run an IPI install using the e2e-vsphere-ovn-dualstack-primaryv6 CI job (vSphere platform, OVN, dual-stack with IPv4+IPv6 machineNetwork and serviceNetwork as in the attached install-config.yaml).

2. Allow the installer to generate Cluster API assets and wait for control-plane provisioning.

3. Observe the Cluster API VSphereMachine and Machine objects for the control-plane nodes in the generated artifacts/log-bundle.

Actual results:

Install fails; control-plane machines never become Ready.

Relevant log excerpts:

From build-log.txt:

level=warning msg=Condition Ready has status: "False", reason: "WaitingForIPAllocation", message: ""
level=warning msg=Condition VMProvisioned has status: "False", reason: "WaitingForIPAllocation", message: ""
level=error msg=failed to fetch Cluster: failed to generate asset "Cluster": failed to create cluster: failed to provision control-plane machines within 15m0s

From Cluster API Machine and VSphereMachine YAML (clusterapi_output and log-bundle):

status.conditions[*].reason: WaitingForIPAllocation
status.infrastructureReady: WaitingForIPAllocation

From VSphereMachine spec (example master-2):

spec.virtualMachineCloneSpec.network.devices[0].networkName: /cidatacenter-2/host/cicluster-3/ci-vlan-958-3
spec.virtualMachineCloneSpec.network.devices[0].dhcp4: true
spec.virtualMachineCloneSpec.network.devices[0].dhcp6: false
  

From rendered MachineConfig for dual-stack masters/workers:

kernelArguments: ["ip=dhcp,dhcp6"]

Combined, these show that IPv6 DHCP is disabled at the VSphereVM level while expected by the OS, and the control-plane never receives the required IP allocation.

Expected results:

Install succeeds.

For dual-stack vSphere IPI installs, the generated VSphereMachine network devices should have DHCP6 enabled dhcp6: true when dual-stack ip=dhcp,dhcp6 is configured, so that control-plane and worker nodes obtain both IPv4 and IPv6 addresses and progress beyond WaitingForIPAllocation.

Additional info:

Cluster type: vSphere IPI, OVN, dual-stack primary IPv6 (e2e-vsphere-ovn-dualstack-primaryv6).

Example install-config.yaml networking:

  - networking.machineNetwork:
    - cidr: fd65:a1a8:60ad:958::/64
    - cidr: 10.93.152.0/25
  - networking.clusterNetwork:
    - cidr: fd65:10:128::/56
    - cidr: 10.128.0.0/14
  - networking.serviceNetwork:
    - fd65:172:16::/112
    - 172.30.0.0/16

Generated artifacts show mismatch between:

• VSphereMachine network device config: dhcp4: true, dhcp6: false
• MachineConfig kernel args: ip=dhcp,dhcp6

This likely originates from the installer/CAPV integration where NetworkDeviceSpec for VSphereMachine is constructed without explicitly enabling DHCP6 for dual-stack, causing dhcp6 to default to false in the rendered VSphereVM spec.