<Python> Build a dedicated server for Jupyter Notebook data analysis

TL;DR I tried to reproduce the steps when the remote server (Cent OS 6.8) was made a dedicated server for Jupyter Notebook. It is convenient because it can be used as a dedicated analysis server for teams.

environment

• Cent OS 6.8
• Anaconda (Python 3.5.1)
• Jupyter Notebook (4.1.0)
• Pyenv (20160509-74-ge93ae00)
• Nginx (nginx/1.10.1)

1. Install Pyenv

Install Pyenv so that you can easily switch between multiple Python versions.

curl -L https://raw.githubusercontent.com/yyuu/pyenv-installer/master/bin/pyenv-installer | bash

pyenv -vWhen the version is displayed with, ok

pyenv 20160509-74-ge93ae00

2. Install Anaconda with Pyenv

Anaconda is a distribution with a lot of data analysis analysis libraries. Jupyter is also in this.

pyenv update
pyenv install -l //Check the latest version of anaconda
pyenv install anaconda3-4.1.0 //Latest version installation
pyenv global anaconda3-4.1.0 //Default Python switches to anaconda

It takes some time to install, but it's OK if the version is displayed with `jupyter --version`

4.1.0

3. Install Nginx

Consider using one IP for multiple services, and use Nginx.

First, register repo so that you can get the latest Nginx

/etc/yum.repos.d/nginx.repo

 [nginx]
 name=nginx repo
 baseurl=http://nginx.org/packages/centos/6/$basearch/
 gpgcheck=0
 enabled=1

Update yum and install Nginx

sudo yum update
sudo yum install nginx

nginx -vWhen the version is displayed with, ok

nginx version: nginx/1.10.1

Include Nginx's Jupyter-specific configuration file jupyter.conf before creating it

/etc/nginx/nginx.conf

http {
  include /etc/nginx/conf.d/jupyter.conf;
}

Create jupyter.conf. Listen to 9999.

/etc/nginx/conf.d/jupyter.conf

location / {
  proxy_http_version 1.1;
  proxy_set_header Upgrade $http_upgrade;
  proxy_set_header Connection "upgrade";
  proxy_set_header Origin "";
  proxy_pass http://127.0.0.1:9999;
}

Start Nginx

sudo service nginx start

4. Jupyter settings

Create a Jupyter config file

jupyter notebook --generate-config

jupyter notebookThe command launches the browser by default. Set the default port used by Notebook to 9999 while preventing it from booting.

.jupyter/jupyter_notebook_config.py

c.NotebookApp.open_browser = False
c.NotebookApp.port = 9999 

5. Finally Jupyter starts!

That's all you have to enter!

jupyter notebook

Access the IP of the server, and when the screen below appears, startup is complete. The files and directories of the directory where the command was entered are displayed.

Immediately press "New" to create a new file

As the name of Notebook suggests, you can write code as a trial writing, and you can execute it line by line, like one line = one set of code. In my case, I imported all the required libraries in the first line so that autocomplete works.

5. Easy ssh connection to Jupyter server

After setting the server port and user, it is troublesome to enter a long ssh command on the client side to connect to the remote, so make it possible to connect with `ssh jupyter`.

~/.ssh/config

Host jupyter
    HostName        IP_ADDRESS
    Port            SSH_PORT
    IdentityFile    SECRET_KEY
    User            USERNAME