Issue with Network Load Balancer TCP Connections

Hello,

I am reaching out to report an issue we are encountering with the Network Load Balancer (NLB) service.

We have a TCP server hosted within an EC2 service in an ECS cluster. It is linked to the NLB via a target group of type IP address. The target group has a health check in TCP to the same port.

We have observed that the devices are experiencing difficulties in sending their requests to the server when passing through the NLB with this configuration.

To diagnose the problem, we conducted several TCP connection tests:

1 - Initial condition: Connection using the same load balancer. Results : The connection is unreliable, and the device fails to establish a connection until after approximately 10 attempts.

2 - Direct connection via the public IP address of the internet box to a local TCP server. Results : The connection is established instantaneously upon the first attempt by the device.

3 - Connection via a new Network Load Balancer created towards a server running in Docker on an EC2 instance. The target group is the EC2 instance, and the health check is in TCP to the same port as the server. Results : The connection is unreliable, and the tag fails to connect until after approximately 10 attempts.

4 - Connection via the new Network Load Balancer created towards a server in Docker on an EC2 instance. The target group is the EC2 instance, and the health check is in TCP to a different port that pings a container Docker different from the TCP server. Results : The connection is established instantaneously upon the first attempt by the device.

Some additional notes:

• We have attempted to modify the values of the health check Timeout and Interval parameters, but this has had no effect on the tests.
• The security groups are correctly configured.
• The target groups in these test setups are properly registered and marked as healthy at the time of testing.

Any guidance or suggestions you can provide would be greatly appreciated.

Thank you!