I'm not sure whether this is a bug or freature request.
Currently the ping operation that's supposed to fetch topology to the hot rod client simply fails when it tries to fetch from a non-functional server in the server list.
It should retry on failure, similar to other cache operations.
The documented purpose of the
infinispan.client.hotrod.ping_on_startup parameter would be something in the line with:
finds the first working server in the server list and obtains the topology information from it.