How to run Jupyter Notebook in server which is at multi hop distance?

This post is in continuation of my previous post “How to set up data science work-space in Ubuntu and Python?”. Today all the budding data scientists are using Jupyter Notebook more frequently. But very often the situation comes where you login to your work-machine from your local machine. Generally, the local machine is your personal computer or laptop. Now all you want to run your notebook in work-machine and access it via the browser in your local machine. We will use ssh tunnelling (link1, link2) in this case.

1-Hop SSH Tunnelling

In the image Host-A is your local machine and Host-B is the work machine. We will do ssh tunneling with port forwarding here. It means while doing ssh we will bind a port, e.g. port-A, of Host-A to a port, e.g. port-B of Host-B. Thus we create a ssh tunnel between these two hosts. So any application running in Host-B:Port-B can be read by an application running at Host-A:Port-A. Next we will run the Jupyter Notebook in Host B at Port-B. Finally in Host A start your browser and go to localhost:port-A. Due to the ssh tunneling with port forwarding the browser in Host-A will be able to read the Jupyter Notebook running in the Host-B. Here are the steps:

Open a terminal in Host-A and run the following:

ssh -L<port_A>:localhost:<port_B> user@Host-B

Next login to Host-B and start the Jupyter Notebook as follows:

jupyter notebook --no-browser --port=<port-B>

Remember, give port number without the bracket<>. Next start the browser in Host-A

2-Hop SSH Tunnelling

Here the scenario is a little complicated than the previous case. Some time we login to our work machine from our local machine via a login host machine. Therefore, we need to create tunnel such that we can connect these three machine. In the above image Host-A is your local machine and now Host-C is the work machine. Where as Host-B is the login host machine. Like the earlier method, we will do ssh tunneling with port forwarding here also with a tweak. Finally we will run the Jupyter Notebook in Host-C at Port-C. And then in Host A start your browser and go to localhost:port-A. Due to the ssh tunneling with port forwarding the browser in Host-A will be able to read the Jupyter Notebook running in the Host-C. Here are the steps.

Simple Steps:

1. First open terminal in Host-A and do ssh from Host-Ato Host-B

ssh -f user@Host-B -L <Port-A>:localhost:<Port-B> -N

2. Now do normal ssh to Host-B and then from there create ssh tunnel to Host-C similarly

ssh -f user@Host-C -L <Port-B>:localhost:<Port-C> -N

3. Now do normal ssh to Host-C and start jupyter notebook in Port-C

4. In Host-A open browser and go to localhost:<Port-A>

and then put the jupyter notebook token from Host-C.

Compact Steps

You can combine the steps 1 and 2 as follows:

ssh -L<Port-A>:localhost:<Port-B> user@Host-B -t ssh -L<Port-B>:localhost:<Port-C> user@Host-C

And then follow steps 3 and 4.

For further understanding you can read these posts: link1, link2.

Happy Coding :)