$ sudo pip install rk

$ sudo pip install git+git://github.com/korniichuk/rk#egg=rk

$ sudo pip install -U rk

or:

$ sudo pip install --upgrade rk

$ sudo pip uninstall rk

$ git clone git://github.com/korniichuk/rk.git
$ cd rk
$ sudo pip install .

First, make sure that you can login to a remote machine without entering password. The most basic form of the command is:

$ ssh REMOTE_HOST

If your username is different on a remote machine, you can specify it by using this syntax:

$ ssh REMOTE_USERNAME@REMOTE_HOST

Example:

$ ssh albert@192.168.0.1

Second, install a template of a remote jupyter kernel to kernels location:

$ rk install-template

Third, change the kernel.json file:

$ sudo gedit /usr/local/share/jupyter/kernels/template/kernel.json

The kernel.json file looks like this:

{
 "argv": [
  "rkscript",
  "python",
  "{connection_file}",
  "remote_username@remote_host"
 ],
 "display_name": "Template",
 "language": "python"
}

For a python2 remote jupyter kernel just change remote_username@remote_host. For example from remote_username@remote_host to albert@192.168.0.1.

Fourth, launch jupyter notebook and check your new remote juputer kernel:

$ jupyter notebook

or:

$ ipython notebook

Choose: Files -> New -> Template.

The standard output for –help:

$ rk -h

or:

$ rk --help

For information on using subcommand "SUBCOMMAND", do:

$ rk SUBCOMMAND -h

or:

$ rk SUBCOMMAND --help

Example:

$ rk install -h

The standard output for –version:

$ rk -v

or:

$ rk --version

Open kernels.json file:

$ sudo gedit /usr/local/lib/python2.7/dist-packages/rk/config/kernels.json

The kernels.json file looks like this:

{
 "template": {
  "display_name": "Template",
  "interpreter": "python",
  "language": "python",
  "remote_host": "remote_username@remote_host"
 }
}

Where:

• template -- the name of a remote jupyter kernel,
  • display_name -- a kernel’s name as it should be displayed in the UI. Unlike the kernel name used in the API, this can contain arbitrary unicode characters [1],
  • interpreter -- an entry point or an absolute path to language interpreter on a remote machine,
  • language -- a name of the language of a kernel. When loading notebooks, if no matching kernelspec key (may differ across machines) is found, a kernel with a matching language will be used. This allows a notebook written on any python or julia kernel to be properly associated with the user's python or julia kernel, even if they aren’t listed under the same name as the author’s [1],
  • remote_host -- just a remote host or, if your username is different on a remote machine, use this syntax: remote username AT remote host.

Change kernels.json file and add info about your remote jupyter kernels, for example like this:

{
 "albert2": {
  "display_name": "Albert Python 2",
  "interpreter": "python2",
  "language": "python",
  "remote_host": "albert@192.168.0.1"
 },
 "albert3": {
  "display_name": "Albert Python 3",
  "interpreter": "python3",
  "language": "python",
  "remote_host": "albert@192.168.0.1"
 }
}

Where:

• albert2, albert3 -- the names of a remote jupyter kernels,
  • Albert Python 2, Albert Python 3 -- the display names for the UI,
  • python2, python3 -- entry points on a remote machine,
  • python -- the name of the language of a remote jupyter kernel,
  • albert -- the remote username on a remote machine, not similar with a username on a local machine,
  • 192.168.0.1 -- the remote host.

Jupyter support the system and the user kernels locations:

The default kernels location in the rk: /usr/local/share/jupyter/kernels.

Change the default kernels location:

$ sudo gedit /usr/local/lib/python2.7/dist-packages/rk/config/rk.ini

$ rk list

$ rk install KERNEL_NAME [KERNEL_NAME ...]

Where:

• KERNEL_NAME -- a name of a remote jupyter kernel in the kernels dict kernels.json.

Example:

$ rk install albert2
$ rk install albert2 albert3

$ rk install-template

$ rk install-all

$ rk uninstall KERNEL_NAME [KERNEL_NAME ...]

Where:

• KERNEL_NAME -- a name of installed remote jupyter kernel.

Example:

$ rk uninstall albert2
$ rk uninstall albert2 albert3

$ rk uninstall-template

$ rk uninstall-all

$ rk ssh

If you are familiar with ssh-keygen, ssh-copy-id and ssh-add, this code also setup SSH for auto login without a password [2]:

$ ssh-keygen -t rsa -b 4096 -N '' -f ~/.ssh/id_rsa
$ ssh-copy-id REMOTE_HOST
$ eval "$(ssh-agent -s)"
$ ssh-add ~/.ssh/id_rsa

The default log files location in the rk: /tmp/rk/log. The name of rk log file, for working remote jupyter kernel, look like this: bree@192.168.0.1_1879-03-14_11.30.00.txt. And the log file looks like this:

date: 1879-03-14 Friday
time: 11:30:00

usernames: bree<->albert
remote host: 192.168.0.1

stdin ports: 37654<->58933
hb ports: 53538<->59782
iopub ports: 45330<->51989
shell ports: 36523<->36107
control ports: 50090<->53633

pids: 16965<->20944

The paramiko log file is available in a local connection file directory. The name of paramiko log file, for working remote jupyter kernel, look like this: paramiko-843664c7-798d-4a9e-979c-22d0dc4a6bd5.txt.

• added
• corrected
• removed

• bug in the rk and in the rkscript: an initial component of ~ or ~user is not replaced in a paths.
• bug in the rk: a superuser (root) privileges required for the user kernels location ~/.ipython/kernels.
• setup SSH for auto login without a password with a "ssh" subcommand.
• error in the rkscript: list index out of range.
• info about working remote jupyter kernel in rk log file.
• paramiko log file in a local connection file dir.
• error in the rkscript: no handlers could be found for logger "paramiko.transport".
• local port forwarding in the rkscript via paramiko, not via pexpect.

• uninstall all jupyter kernels from kernels location with a "uninstall-all" subcommand.
• uninstall a remote jupyter kernel/kernels from kernels location with a "uninstall" subcommand.
• install a remote jupyter kernel/kernels from kernels dict to kernels location with a "install" subcommand.
• install all remote jupyter kernels from kernels dict to kernels location with a "install-all" subcommand.
• show list of remote jupyter kernels from kernels dict with a "list" subcommand.

Footnotes