The following user guide is for customizing and enhancing the scripts to suit your needs. We cover the following:
- How to password protect your instance
- How to customize the packages that are pre-installed by your instance
Jupyter notebooks can be password protected by replacing a line of code in the deploy-notebook/tasks/main.yml script from:
--NotebookApp.token=''
To:
--NotebookApp.password='PASSWORD_GOES_HERE'
Docker has provided its users with a series of pre-configured 'Docker Stacks', which are pre-built configuration instructions that specify which packages will be installed.
To change the packages that your environment will install, you can access the list of available Docker Stacks here. Replacing the Docker Stack within the file: config/roles/deploy_scipy_notebook/tasks/main.yml with the desired Docker Stack.
jupyter/scipy-notebook
to
jupyter/whatever
Customizing which files are installed each time you set up an instance requires the following
- a free account with Docker
- a github account to host your dockerfile
First log into your github and create yourself a new repository where you'll store all your DockerFiles. I made a folder where I'll store the dockerfile for this example, called mike-custom. We'll put the dockerfile in this folder. I have an example here from my own account of what this looks like
Feel free to fork this, or start with your own. I started with a dockerfile from the Docker Stack called scipy-notebook. If you open the dockerfile, you'll find a an immediately recognizable list of packages like pandas, matplotlib, and scipy. Lets say I wanted to add xgboost to my package list thats automatically downloaded. I'd either just add 'xgboost' \ to the list, or if I wanted a more specific version, I'd look it up on the python package index. In which case, I'd specify 'xgboost=0.6*' \ The star after 0.6 connotes that you're okay with any version change off 0.6 like 0.61 or 0.6a2.
After you push this to your GitHub account, take note of the directory. You should know your repository name, any folder you created, and the name of the dockerfile in your folder.
Next, you'll go to Docker and create and account there if you don't have one. After you create your Docker account, go and click 'Create Repository' to get started. Give your repository a name [not necessarily the same as your github repo] and write a description.
Now connect to your repository to your GitHub account and click 'Create'. On the next page, navigate to the 'Build' tab, and here you will specify your GitHub repo's subpage.
After you click build, you should receive a success message. At this point, you'll modify the notebook run code described in #2 to direct from Jupyter's stacks to your Docker account.
jupyter/scipy-notebook
to
urchinlabs/mike [Docker Repo / repo name]