I have deployed cluster of two Ignite 2.5.0 servers on two hosts (OpenStack cloud). They see each other using ipFinder TcpDiscoveryMulticastIpFinder (Topology snapshot [ver=14, servers=2, clients=0, CPUs=2, ...). Only one host has a public IP (floating IP), let's call it A, the other has only private IP, B, accessible from host A only. In my client (Eclipse IDE), in the config, I have set the public IP of A to connect to with TcpDiscoveryVmIpFinder.addresses. (In the config of A, public and private IPs in addressResolver and localAddress are correctly set).
When sending a broadcast() to this cluster, nothing happens for minutes. Killing B, broadcast works (on A).
How can I access "private" compute nodes of the cluster from a client outside? I saw signs of a "router" (bin/igniterouter.sh, config/router/default-router.xml) - maybe deprecated?, but not much info on how to use it, or any other way to solve this problem.