Working with jupyter server remotely

I normally set-up my desktop as a server and my laptop as a client. Jupyter notebook can be run in a remote server which can be edited locally in a client’s browser. In this way, the computing resources of my desktop is utlized with the convenience of working on a laptop.

We will do this by opening an SSH tunnel. This tunnel will forward the port used by the Jupyter kernel running on remote server to a port on the local machine, where it can be accessed in a browser just like running Jupyter in your local computer.

Setting up jupyter in a remote server

In your remote server, launch jupyter notebook without browser

[remote-server@user] $ jupyter notebook --no-browser

Using jupyter notebook and viewing it on your local browser

  1. From your local computer, connect to the Jupyter kernel running on s remote server via ssh tunnel with the format: ssh -L remote_host:remote_port:localhost:locat_port
$ ssh -L localhost:8888:localhost:8889 <remote-server>

where 8888 is your local browser’s (unused) port number and 8889 is remote’s port number currently used by Jupyter.

If you see a message saying:

> (bind) port already in use

That means 8888 is not available (may be you have other jupyter kernel running locally), so you can try 8889, 8890, and so on. I’ll use 9999:

$ portnumer = 9999
$ ssh -L localhost:$portnumber:localhost:8889 <remote-server>

Note that you can check which port number is being used:

$ jupyter notebook list
  1. To view remote jupyter server using your local browser, go to:

localhost:9999

  1. If asked for a token (usually for first time users), run the command in ssh session. 
    [remote-server@user]$ jupyter notebook list

Example output:

Currently running servers:
http://localhost:9999/?token=6bb0d3427636d0f0802fd1b6eeacaa9f84e9326cfe7fb1c7 :: /home/usr

Copy 6bb0d3427636d0f0802fd1b6eeacaa9f84e9326cfe7fb1c7 and paste it in your browser.

  1. To avoid ssh disconnection, run htop in ssh session and leave it as it is: 
    [remote-server@user] htop
  2. To quit Jupyter notebook, first check the process id: 
    $ ps aux | grep localhost:9999

Sample output: 23295 0.0 0.0 50064 6608 pts/2 S+ 16:32 0:00 ssh -L localhost:9999:localhost:8889

Then kill the process in your local pc:

$ kill 23295

Finally, you can disconnect from ssh session in the usual way. First, quit htop by pressing key and then type in the remote server terminal.

For convenience, I add the following in my ~/.alias sourced in ~/.bashrc:

portnumber=9999
alias ssh_jup="echo 'using local port number=' $portnumber && ssh -L localhost:$portnumber:localhost:8889 <remote-server>"