Running Jupyter Notebooks on Remote ServersTobias Skovgaard JepsenBlockedUnblockFollowFollowingMar 4Jupyter Notebook are a staple tool in many data scientists toolkit.
Jupyter Notebook makes it easy to perform and visualize data analysis, prototyping models, and running experiments by shortening the feedback loop from coding to seeing results.
In many cases, running a Jupyter notebook on a laptop or work station is sufficient.
However, if you’re working with large datasets, doing computationally expensive data processing, or learning complex models you probably need something more powerful than a laptop extra power.
Maybe you’re running graph convolutional networks on large graphs or doing machine translation with recurrent neural networks on large text corpora and need some more CPU cores, RAM, or a couple of GPUs.
Luckily, you may be in the situation where you have these resources available on a remote server!If your remote server has a Graphical User Interface (GUI), you are in luck.
You can use remote desktop software to access the remote server and otherwise use Jupyter Notebook as your normally would on your laptop.
However, many servers do not have a GUI.
If you are in this situation, you can set up your experiment by writing a Python script on your laptop, run it on a small subset of your data to verify that it can run, copy it to a remote server, and execute it from the command line.
You could even set up the experiment in a notebook and export the notebook to a script using jupyter nbconvert –to script your_notebook.
ipynb.
Although this workflow certainly allows you to run your code on the remote server, you can no longer use Jupyter Notebook to, e.
g.
, experiment with your models and visualize your results interactively.
What a shame!In this post I will show you how to run a Jupyter notebook on a remote server and how to access it on your laptop.
I will also show how to setup two bash commands to make the whole process easier.
Starting the Remote Notebook ServerWe will use the Secure Shell Protocol (SSH) to start the Jupyter Notebook server on the remote server.
SSH allows us to send commands to the remote server.
The basic syntax is as follows:ssh username:password@remote_server_ip commandThe exact command you should send depends a little on your context.
In my case, I share a remote server with other people have therefore not installed Jupyter in the shared environment.
The first step for me is therefore to go to my project folder, activate the virtual environment, and start the notebook server.
In particular, I would like to execute the following threebash commands on the remote server:cd project_folder.
virtual_environment/bin/activatejupyter notebook –no-browser –port=8889I execute the jupyter notebook command with the –no-browser flag to start the Jupyter notebook with launching a browser since the remote server cannot display a browser if it doesn’t have a GUI.
I also change the port from the default port 8888 to port 8889 using the –port=8889 flag.
This is a personal preference; having local and remote notebooks on different ports to make it easier to see where my code is running.
To execute commands on the remote server, we run the combined commandnohup ssh -f username:password@remote_server_ip "cd project_folder; .
virtual_environment/bin/activate; jupyter notebook –no-browser –port=8889"Note that I have one-lined the three commands and separated them using ; instead of line breaks.
Executing this command will start the Jupyter Notebook server on port 8889 and let it run in the background.
Finally, I have added the -f flag to the ssh command to push the process to the background and prepended the nohup command to silence all output from the process so you can continue using the terminal window.
You can read more about thenohup command here.
Accessing Your NotebookYou can now access the notebook typing in the urlremote_server_ip:8889This command requires you to memorise the IP address or to bookmark the web page.
However, we can make it just as easy to access the remote notebook as if it were a local notebook by using port forwarding:nohup ssh -N -f -L localhost:8889:localhost:8889 username:password@remote_server_ipThe -N flag tells ssh that no remote commands will be executed.
At this point we do not need to execute any remote commands.
The -f flag pushes the ssh process to the background as mentioned previously.
Finally, the -L flag specifies the port forwarding configuration using the syntax local_server:local_port:remote_server:remote_port.
The configuration specifies that all requests sent to port 8889 on the local machine, e.
g.
, your laptop, to port 8889 on the remote machine at username:password@remote_server_ip.
As before, the nohup command has been prepended to silence the output.
The effect the above command is that you can now access the remote Jupyter Notebook server in your browser atlocalhost:8889as if you ran the notebook locally.
Stopping the Remote Notebook ServerIn principle, you can let the notebook server run indefinitely on the remote server (barring restarts or crashes), but you may need to stop the server, for instance to upgrade your version of jupyter.
If you need to stop it there are two ways to do so: through the browser or through the command line.
Through the Browser WindowIn the recent versions of the Jupyter Notebook, you can find a Quit button at the top right of the browser window as indicated by the arrow in the image below.
If you press it, you will have to relaunch the server again using the start-up command we saw previously.
The Quit ButtonThrough the Command LineIf you are unable to upgrade to a newer version of Jupyter that has the Quit button or simply prefer working through a terminal, you can also stop the server from the command line.
Jupyter has a shell command to stop notebooks:jupyter notebook stop 8889where 8889 is the port number.
You can execute it on the remote server using the commandssh username:password@remote_server_ip "jupyter notebook stop 8889"Unfortunately, this command is currently bugged, but I have included it here in the hopes that it will work in the future.
However, as a work-around you can instead kill the jupyter process instead using the command:ssh username:password@remote_server_ip "pkill -u username jupyter"where the -u username indicates that only jupyter processes started by username should be killed.
The drawback of doing this is that you will shutdown all of your notebook servers if you have more than one running at a time.
Finally, you can of course manage the servers manually by logging on to the remote server, starting the notebook server, and keeping the terminal window open.
This allows you to shutdown the notebook server using the usual CTRL+C keyboard command.
Smoothening Your WorkflowRemembering all these commands can be quite cumbersome.
Thankfully, we can make life easier by creating bash aliases for each of the commands.
Add the following lines to your ~/.
bashrc file:Load the commands by typing source .
bashrc in your terminal.
You can now use the commands remote_notebook_start and remote_notebook_stop in your terminal to respectively start the remote notebook server (and setup port forwarding) and shut it down.
Wrap-UpIn this post I have shown you how to start, access, and stop Jupyter notebooks on remote servers using bash commands and shown how to create bash aliases to make it easy to do so.
I hope that these commands can improve your data science productivity by nearly seamlessly allowing you to reap the benefits of both Jupyter notebook and any computing resources you have available on remote servers.
Liked what you read?.Consider following me on Twitter where I share papers, videos, and articles related to the practice, theory, and ethics of data science and machine learning that I find interesting, in addition to my own posts.
For professional inquiries, please contact me on LinkedIn or by direct message on Twitter.
.