Loading data from multiple remote servers in tensorboard

This is more a feature question rather than a bug. I hope others will find it useful too.

I am running tensorboard on a remote server, server_A. I was wondering if I somehow can get tensorboard to also look at the event files stored in the runs directory on a different server, server_B. I have started with the command below:

tensorboard --logdir=path/to/logs/1,server_B:/path/to/logs/2

That did not work so I simplified a bit and tried to get tensorboard to "look" at the event files on the remote alone. I tried the following:

tensorboard --logdir=/path/to/logs/2 --host=112.11.15.72 --port=8888

where the ip in host is the private IP address of server_B, for which I configured an inbound Custom TCP connection on that port. The error I get is:

E1018 11:44:05.649294 MainThread program.py:201] TensorBoard could not bind to unavailable address 112.11.15.72
E1018 11:44:05.649293 140128884926208 program.py:201] TensorBoard could not bind to unavailable address 112.11.15.72
ERROR: TensorBoard could not bind to unavailable address 112.11.15.72

Is it possible to achieve this?

Start by reproducing the two command-line examples using tensorboard --logdir and --host, noting the remote-server setup and bind error shown in the issue. Determine whether remote event-file locations are intended to be supported; done means either a documented working approach or a defined implementation path with coverage for the requested behavior.