I have a manager node with a fixed ip address, and when I join the swarm with a worker on EC2, I can see in the docker logs that the manager tries to connect to the internal ip address of the EC2 instance. According to docker, this should not happen and the manager node should be able to determine the public ip address of the joining node.
When the worker joins the swarm with the --advertise-addr="public ip" flag set, everything works correctly, but this is not an option due to dynamic ip. Thus when the flag isn't specified, the worker tries to determine the address from the interfaces, find the local one, and advertise it.
How can I prevent the worker node to advertise its local address, and let the manager figure out the node ip address itself ?
