Tensorboard w/ Jupyter server Proxy

Hi I'm trying to use Jupyter server proxy extension to run Tensorboard and access it directly! I tested with /proxy/port and it works great, however I wanted to add the possibility to start it using https://jupyter-server-proxy.readthedocs.io/en/latest/server-process.html#specifying-config-via-traitlets (to be able to use simply the new tensorboard button and not have to start a terminal each time).

I created a file '''jupyter_notebook_config.py''' copied to /root/.jupyter/jupyter_notebook_config.py with the following content :

c.ServerProxy.servers = {
  'tensorboard': {
    'command': ['tensorboard', '--logdir', '/tensorboard' , '--port', '{port}']
  }
}

The first issue is that I dont see tensorboard option in the new toolbar but its not so important...
If i access to /tensorboard/ Tensorboards starts to run (i can see it in htop) but I get a 504 bad gateway after a while.
- If I access /proxy/portNumberGeneratedRandomly it works !

Do you have an idea about whats going on ?

Thank you for your help,

Config: Inside Kubeflow (kubernetes) last version of Jupyter server Proxy and version 1.15 of Tensorboard.

jupyterhub / jupyter-server-proxy