I’m curious about the if we can have multiple Ray client servers in a Ray cluster. If we have a lot of clients connecting to the client server, the server might become a bottleneck. It would be great if we can launch multiple client servers on different nodes to balance the load.
It seems to me that the underlying implementation of the client server actually supports this. So is there any interface to achieve it?
Thanks!
hi @Hanyu
Unfortunately, this is not officially supported at the moment. However, if you are interested you can submit a PR to make that possible. (related code: ray/services.py at master · ray-project/ray · GitHub)