Tensorboard: 404 error when used with a Jupyter proxy

[hadim]

• TensorBoard 1.8.0
• Tensorflow 1.8.0
• Linux Ubuntu 18.04
• Python 3.6
• Chromium 66

I am trying to use a Jupyter proxy called nbserverproxy in order to access TB from the Jupyter server (that saves me one more SSH tunnel with the cluster).

TB interface loads correctly but can't retrieve information from /data/runs, /data/environments, and so forth. Opening the browser console shows a lot of 404 errors.

With nbserverproxy you access TB with the following URL: http://localhost:8888/proxy/6006. Looking at the browser console, TB seems to be looking for http://localhost:8888/proxy/data/runs instead of http://localhost:8888/proxy/6006/data/runs.

I tried a bunch of --path_prefix, --base_url and --host options, I can't get it to work.

I am not sure if it's a technical limitation, a bug from TB or something I am doing wrong.

Here is the associated issue on the nbserverproxy: https://github.com/jupyterhub/nbserverproxy/issues/41

[nfelt]

I think you might be seeing this due to a missing trailing slash in the URL you're using to visit TensorBoard - what happens if you try it with http://localhost:8888/proxy/6006/ including the trailing slash? You might also need to pass --path_prefix=/proxy/6006/ depending on how nbserverproxy works (I'm not familiar with it).

Without the trailing slash, when tensorboard requests the relative path data/runs it's probably dropping the 6006 from /proxy/6006 as though it were the filename and getting /proxy/data/runs instead.

[hadim]

Pretty easy fix indeed!!!

The trailing slash did the trick :-)

[nfelt]

Glad to hear that fixed it!

[canavandl]

thank you! thank you! thank you! @nfelt