Description of problem:

In an OpenShift cluster with an OpenShiftSDN network plugin with egressIP and NMstate operator configured, some conditions exist when the egressIP is deconfigured from the network interface.

This issue has been fixed in multiple versions of OpenShift but the customer is facing the same. Customer is running 4.12.35, which should already have the patch for the issue according to the KCS.

https://access.redhat.com/solutions/6971750

https://issues.redhat.com/browse/OCPBUGS-32713
The bug is 100% reproducible.

The steps for reproducing the issue are:

1. Install a cluster with OpenShiftSDN network plugin.

2. Configure egress IP for a project.

3. Install the NMstate operator.

4. Create a NodeNetworkConfigurationPolicy.

5. Identify on which node the egress IP is present.

6. Restart the nmstate-handler pod running on the identified node.

7. Verify that the egress IP is no longer present.

Restarting the sdn pod related to the identified node will reconfigure the egress IP in the node.

This issue has a high impact since any changes triggered for the NMstate operator will prevent application traffic. For example, in the customer environment, the issue is triggered any time a new node is added to the cluster.

The expectation is that the NMstate operator should not interfere with SDN configuration.