Remote jupyter Sessions With ob-ipython

While ssh'd into the server (and with the appropriate virtual environment running if you need it), enter the command to show the folder where the JSON file will be.

jupyter --runtime-dir

In my case this is what came out.

/run/user/1000/jupyter

Change into the directory where you want the jupyter server to run and start the ipython kernel.

ipython kernel

This is what came out for me.

NOTE: When using the `ipython kernel` entry point, Ctrl-C will not work.

To exit, you will have to explicitly quit this process, by either sending
"quit" from a client, or using Ctrl-\ in UNIX-like environments.

To read more about this, see https://github.com/ipython/ipython/issues/2049


To connect another client to this kernel, use:
    --existing kernel-10181.json

That last line refers to the JSON file that we're going to need on the client side (kernel-10181.json).

Go back to your client (my laptop in this case) and check where your jupyter installation is.

jupyter --runtime-dir

/run/user/1000/jupyter

Both my laptop and desktop had the same location, so it's probably the default. Now change into that directory on the client and copy the file from the server.

cd /run/user/1000/jupyter
scp Hades:/run/user/1000/jupyter/kernel-10181.json .

You can run this next command anywhere on the client where you can run jupyter.

jupyter console --existing kernel-10181.json --ssh Hades

Where kernel-10181.json is the files copied from the server and Hades is the ssh alias for my server (so it would be hades@erebus without the alias).

When you run this command it will open up an ipython prompt that will be connected to the server. This turns out to be more useful than I thought it would be because the prompt that ob-ipython normally opens seems to be broken when I'm connected to the remote client. Maybe there's something else to configure. Anyway, whatever you add to your emacs ob-ipython namespace will be accessible to you in the ipython prompt so you can fiddle with things live in there.

Once you open this it will add another file (kernel-10181-ssh.json in this case) that you can use to connect if you want to open up more than ipython prompt for some reason:

jupyter console --existing kernel-10181-ssh.json

You don't pass in the --ssh flag this time. Note that you're still sharing the same jupyter session so all the variables and stuff will show up in the second console as well.

But, more importantly, this second json file is how we can connect while in emacs.

To make use of the session you need to pass in the name of the ssh kernel file as the name of the ipython session.

#+BEGIN_SRC ipython :session kernel-10181-ssh.json :results none

#+END_SRC

Anything you do in a block with that session ID will send the commands to the remote server to be interpreted.

To clarify:

• kernel-10181.json is the file you copy over from the server
• kernel-10181-ssh.json is created on your client and is what you use in the org-babel header

%matplotlib inline
seaborn.set(style="whitegrid",
            rc={"axes.grid": False,
                "font.family": ["sans-serif"],
                "font.sans-serif": ["Latin Modern Sans", "Lato"],
                "figure.figsize": (13, 13)},
            font_scale=1)

Load the brain networks example dataset

data = seaborn.load_dataset("brain_networks", header=[0, 1, 2], index_col=0)

Select a subset of the networks.

used_networks = [1, 5, 6, 7, 8, 12, 13, 17]
used_columns = (data.columns.get_level_values("network")
                .astype(int)
                .isin(used_networks))
data = data.loc[:, used_columns]

Create a categorical palette to identify the networks.

network_palette = seaborn.husl_palette(8, s=.45)
# I have no idea what the network_lut is.
network_lut = dict(zip(map(str, used_networks), network_palette))

Convert the palette to vectors that will be drawn on the side of the matrix.

networks = data.columns.get_level_values("network")
network_colors = pandas.Series(networks, index=data.columns).map(network_lut)

Plot.

cluster = seaborn.clustermap(data.corr(), center=0, cmap="vlag",
                             row_colors=network_colors,
                             col_colors=network_colors,
                             linewidths=.75)
title = cluster.fig.suptitle("Brain Networks")

Besides the ipython shell not working in emacs I also couldn't get it to render images directly in emacs, but they did render on the remote side so I could see the plot in my web-browser.