A customer has a requirement to configure an additional IP to the worker node/pod (by creating a new network interface belonging to a new subnet)
Then the pod with an IP assigned from secondary interface would talk to an external system which belong to the same network.

We attempted several tests and observed that the connection from inside the pod (having secondary interface via NAD) cannot go out side of the node network. 

Here's the sample of NAD configuration we used:
~~~
spec:
  config: '{ "cniVersion": "0.3.1", "type": "ipvlan", "mode": "l3", "master": "ens6", "ipam": { "type": "whereabouts", "range": "10.1.0.128/28", "exclude": ["10.1.0.129/32"]  } }'
~~~

We also tried l2 mode in the NAD but no luck.

OCP 4.14

Always    

1. Configure an OCP cluster on AWS
2. Attached a secondary interface to one or more worker nodes.
3. Configure NAD with IPVLAN in l2/l3 mode
4. Deploy a sample app who's pod gets assigned with an IP via NAD (via k8s.v1.cni.cncf.io/networks: <NAD_name>
5. Verify pod gets the IP
6 Try running ping or curl against external system (outside of OCP) which is in the same network. During reproducer we simply configured a RHEL9 ec2 instance of same network, allowing necessary security groups.
    

- Ping gets stuck without any specific error. (we ensured necessary security groups with inbound/outbound rules are configured)

- Communication should happen from pod to external system which is in the same network as pod.    

We already had some discussion on [forum-ocp-multus](https://redhat-internal.slack.com/archives/CFFSAHWHF/p1738930040270059) but we'd probably need a concrete confirmation to know if we support multus IPVLAN with this kind of scenario on AWS based OCP cluster.