Hairpin connection failed on Service type-LoadBalancer NLB with internal scheme.

The hairpin connection impacts any application which the client and server are hosted in the same node, exposed by a Service type-LoadBalancer NLB (only). The CCM creates a NLB with preserve source IP address attribute enabled by default, and does not provide interface to change it, leading to fail in those scenarios.

The Default router deployed by Cluster Ingress Controller on OCP standalone on AWS and ROSA HCP private deployments.

For more information, see upstream issue[1], and e2e tests[2] reproducing the problem on NLB.

Always

Scenario 1) Standalone Service app:
- Deploy a sample server application sticking to single node
- Expose the server through a Service type-LB NLB with private scheme 
- Run a client/curl to reach the LB endpoint from the same node the server was deployed

Scenario 2) Using OCP Default Router
- Create OCP/ROSA HCP private cluster
- Deploy a sample server application sticking to single node
- Expose the server app through a router 
- Run a client/curl to reach the router endpoint from the same node the server was deployed

Scenario 1) connection timeout for single node
Scenario 2) Eventually connection timeouts depending the number of replicas the router have in the cluster    

hairpin connection works in the private service type-LoadBalancers    

[1] Upstream CCM-AWS Issue: https://github.com/kubernetes/cloud-provider-aws/issues/1160
[2] Upstream e2e tests reproducing the problem: https://github.com/kubernetes/cloud-provider-aws/pull/1161