Remote Kernel v1.1

Remote kernel is a pure-python package that sets up an ssh connection with port forwarding to a remote host, and starts up an IPyKernel on that host. This allows a locally running instance of Jupyter or Spyder to easily run code on a remote server. Moreover, this can be installed as a jupyter 'kernel', allowing you to use this just as easily as you would use a local kernel!

Installation

To install remote_kernel, run:

pip install remote-kernel

Usage

Remote kernel is intended to spawn kernels on remote servers for use in both Jupyter notebooks and Spyder.

Jupyter

Install the remote kernel:

python -m remote_kernel install -t <ssh_host> [Options]

To obtain a list of options for the installation, run:

python -m remote_kernel install --help

Spyder

To use the remote kernel in spyder, a kernel needs to be manually started and then connected to from spyder.

To start the remote kernel:

python -m remote_kernel -t <ssh_host> [Options]

Again, options can be listed using:

python -m remote_kernel --help

Alternatively, it is possible to start the remote kernel from a kernel spec file (i.e. what jupyter uses to start kernels):

python -m remote_kernel from-spec <kernel-name>

Here, <kernel-name> is the name of directory containing the kernel.json kernel spec file. If the name is relative, remote_kernel looks for it relative to the current directory, as well as in the jupyter kernel directories, where jupyter kernels are installed as directories with the kernel name.

If successfull, this will print a message like this:

remote_kernel.start I: Remote Kernel started. To connect another client to this kernel, use:
    --existing kernel-<user>@<host>.json

Copy this name (or just the <user>@<host part) and use it to connect to a local existing kernel in spyder. Even though the kernel is running remotely, and we're using SSH, connect to it as if it were running locally (i.e. don't check the SSH box). N.B. The convenience function with the id expansion also works using the <user>@<host> part, even though it's not a number.

N.B. By default, remote_kernel starts regular ipykernels on the remote server, but this can be overridden using the -c command line option.

Acknowledgements/Requirements

This package relies heaviliy on the following packages

• paramiko, for setting up the ssh connection.
• sshtunnel, for creating the forwarding tunnels for the kernel ports.