JupyterHub/Kubespanwer: found procedure to get working

[tlkh]

Hi everyone, I figured out how to get this working in a JupyterHub/Kubespawner setup:

(assume packages are already installed)

1. Log in
2. Run jupyter tensorboard enable
3. Stop server, login again
4. It works!

Reason: jupyter tensorboard enable cannot be run as part of Dockerfile since the home directory is not mounted until after the container is created. A bit annoying but I'm not yet sure how to work around it.

[consideRatio]

@tlkh thanks for documenting this! I think that if you use the --sys-prefix you can enable it outside of the users home directory. Below you see my output from jupyter serverextension list that shows how it is enabled in two configuration files because I've now run it with and without the --sys-prefx.

config dir: /home/jovyan/.jupyter
    jupyter_tensorboard  enabled
    - Validating...
      jupyter_tensorboard 0.1.8 OK
config dir: /opt/conda/etc/jupyter
    jupyterlab  enabled
    - Validating...
      jupyterlab 0.35.4 OK
    nbdime  enabled
    - Validating...
      nbdime 1.0.3 OK
    nbgitpuller  enabled
    - Validating...
      nbgitpuller 0.6.1 OK
    nbresuse  enabled
    - Validating...
      nbresuse  OK
    nbserverproxy  enabled
    - Validating...
      nbserverproxy  OK
    nbrsessionproxy  enabled
    - Validating...
      nbrsessionproxy  OK
    jupyterlab_iframe.extension  enabled
    - Validating...
      jupyterlab_iframe.extension  OK
    jupyter_tensorboard  enabled
    - Validating...
      jupyter_tensorboard 0.1.8 OK

[gsnidero]

Hi,

I also find that with the --sys-prefx it still does not work. Every user in jhub need to run jupyter tensorboard enable.