Running Jupyter from a remote server

notes
jupyter
How to run Jupyter on a remote server and access it locally

TL;DR

SSH into the remote server and run Jupyter:

ssh [user@remote]
jupyter lab --no-browser --port=[port_in_remote]

Note that the output of the last command should include a token. We will need it when accessing the remote server from the local machine (next step).

On the local machine, create an SSH tunnel to forward the remote port to your local machine:

ssh -NfL localhost:[port_in_local]:localhost:[port_in_remote] [user@remote]

Finally, on your local browser, access http://localhost:[port_in_local]. The token produced in step one may be necessary to access the Jupyter server.

Breakdown

  1. We run Jupyter in the remote machine using --no-browser so that no browser is opened in the remote machine.
  2. Any available port can be specified to avoid port collisions.
  3. The -Nf options for the SSH tunnel run the SSH session in the background (-f) without executing remote commands (-N).
  4. The item localhost:[port_in_local]:localhost:[port_in_remote] can be split in two parts. The first localhost:[port_in_local] specifies the local address (the address on the local machine where the port is being opened), in this case localhost, indicating that only the local machine can access that port, and the local port, [port_in_local], where the tunnel will listen. The second localhost:[port_in_remote] indicates the address on the remote machine to connect to, here localhost, meaning that SSH will connect to a service running on the local machine, and [port_in_remote], which specifies the port in the remote machine to which the tunnel connects.
  5. In order to access the remote Jupyter server, we access the local port of the SSH tunnel via http://localhost:[port_in_local].

Resources