Skip to content
This repository has been archived by the owner on Dec 8, 2022. It is now read-only.

jupyter/nb2kg

Repository files navigation

Jupyter Notebook Extension to Kernel Gateway (nb2kg)

Note: this repository is now part of Jupyter Notebook 6.0!

As of Jupyter Notebook 6.0 the functionality in this repository is now part of the notebook server, and can be accessed via the following command line flag:

--gateway-url=<Unicode> (GatewayClient.url)
    Default: None
    The url of the Kernel or Enterprise Gateway server where
    kernel specifications are defined and kernel management
    takes place. If defined, this Notebook server acts as a
    proxy for all kernel management and kernel
    specification retrieval. (JUPYTER_GATEWAY_URL env var)

This means that there is no need to install this package (unless using notebook server < 6.0).

Overview

nb2kg is a Jupyter Notebook server extension that enables the Notebook server to use remote kernels hosted by a Jupyter "Gateway" (i.e., Kernel Gateway or Enterprise Gateway).

The extension may be useful in cases where you want a local Notebook server to connect to a kernel that executes code on a compute cluster in the cloud, perhaps near big data (e.g., the kernel is a driver program running on an Apache Spark cluster).

The extension overrides the /api/kernels/* and /api/kernelspecs request handlers of the Notebook server, and proxies all requests for these resources to the Gateway. When you enable the extension, all kernels run on the configured Gateway instead of on the Notebook server host (although kernels can also be remoted from Enterprise Gateway servers).

Jupyter remote kernels

The nb2kg extension communicates with the Gateway using standard HTTP and web socket protocols. This differs from other remote kernel projects, such as remote_ikernel and rk, which rely on SSH or other mechanisms to communicate with kernels.

The extension requires Jupyter Notebook 4.2 or later, with support for server extensions.

jupyter serverextension list

Install

To install the released nb2kg extension in an existing Notebook server environment:

pip install nb2kg

To install the latest nb2kg extension in an existing Notebook server environment:

pip install "git+https://github.com/jupyter-incubator/nb2kg.git#egg=nb2kg"

Once the package has been installed, it must then be registered as an extension:

jupyter serverextension enable --py nb2kg --sys-prefix

Run Notebook server

When you run the Notebook server with the nb2kg extension enabled, you must set the KG_URL environment variable to the URL of the kernel or enterprise gateway and you must override the default kernel, kernel spec, and session managers:

export KG_URL=http://kg-host:port
jupyter notebook \
  --NotebookApp.session_manager_class=nb2kg.managers.SessionManager \
  --NotebookApp.kernel_manager_class=nb2kg.managers.RemoteKernelManager \
  --NotebookApp.kernel_spec_manager_class=nb2kg.managers.RemoteKernelSpecManager 

Try It

You can use the included Dockerfiles to build and run a Notebook server with nb2kg enabled and a Kernel Gateway in separate Docker containers.

git clone https://github.com/jupyter-incubator/nb2kg.git
cd nb2kg

Build Notebook server and Kernel Gateway Docker images.

docker-compose build

Run the containers.

docker-compose up -d

Launch a web browser to the Notebook server. On Mac OS X:

open http://my.docker.host:9888

Develop

If you want to modify the extension, you can develop it within your Jupyter Notebook dev environment.

Clone this repo.

git clone https://github.com/jupyter-incubator/nb2kg.git
cd nb2kg

Install and enable the extension.

make install

Run the Jupyter Notebook server.

make dev

Uninstall

To uninstall the nb2kg extension:

jupyter serverextension disable --py nb2kg --sys-prefix
pip uninstall -y nb2kg

Caveats

  • When you enable the extension, all kernels run on (are managed by) the configured Gateway, instead of on the Notebook server host. The extension does not support local kernels.
  • When you enable the extension, notebooks and other files reside on the Notebook server, which means that remote kernels may not have access to them.
  • If your kernel gateway instance is using a self-signed certificate in your development environment, you can turn off certificate validation by setting VALIDATE_KG_CERT=no in your environment before starting the notebook server.