Peter, I have experience of using kernel-gateway in the http-mode. The 
ability to create endpoints with one comment is super-cool, however I felt 
the case was self-limiting as one had to specify one notebook as seed_uri 
per kernel-gateway process. 

1. I feel http-mode belongs to a separate repo as the use case is very 
different from kernel-gateway.
2. kg-http-mode should really support a mode where you can specify the 
seed_uri via an API - while the service is running. This way you could do a 
requests.get to a cell of a notebook running in Jupyter notebook server - 
for each of running kernels, not just one.
3. nb2kg is even more powerful extension that allows offloading 
computational power to a dedicated box. It is the key component that 
ensures the notebooks and the dashboards server reference exactly the same 
environment. An example 
in 
https://github.com/jupyter-incubator/dashboards_setup/blob/master/docker_deploy/Dockerfile.notebook
 
is suboptimal as it installs required dependencies twice - once for the 
kernels container, and second time for the Jupyter notebook server. What if 
your environment a bit more complex and makes use of environment variables 
too? I think a better use case of making the notebook server and the 
gateway referencing the same kernel environment is presented here: 
https://github.com/rs2/pyconuk-2016-jupyter-workshop/blob/master/docker/Dockerfile.notebook

nb2kg really needs to be available on pypi/conda-forge. Happy to give a 
hand, I think it's a really powerful project.


On Thursday, November 10, 2016 at 2:52:08 PM UTC, Peter Parente wrote:
>
> Hi Fabian
>
> Is posible publish many services of the many notebook using jupyter kernel 
>> gateway for the same port all for example I have 3 notebook and for each 
>> notebook there are 3 services Can I publish all for the same port?.
>>
>
> You can only specify one notebook at present. It might be possible to add 
> support for the case you mention. It'll take a bit of thinking about the 
> design: a separate kernel pool per notebook, evaluating all of the 
> notebooks in a single pool, specifying multiple notebooks/specs on the 
> command line, handling conflicts, etc. Alternatively, it might turn out to 
> be simpler to put a proxy in front of separate servers or use a web 
> framework instead of a notebook for the more complex cases.
>
> If you have a strong use case for this support, please open a PR or issue 
> on the kernel gateway repo.
>
> Cheers,
> Pete
>

-- 
You received this message because you are subscribed to the Google Groups 
"Project Jupyter" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/jupyter/89269c8d-fb43-4c6e-83c7-982b37e0b875%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to