Using notebooks from the Web UI
Getting started
There are two different ways to get started with notebooks:
- CLI
- Web
From your project directory, you should run:
conveyor notebook create --env ENV
When asked how you want to access your notebook, you should select the web-ui option.
This command will create a Docker container with a JupyterLab environment with your source code and notebooks if any exist. The created container is linked to your personal credentials and will not visible to other users. The command will display the URL of the newly created JupyterLab environment.
The first time you create a notebook, it will ask questions in order to configure your notebook and store this configuration in the notebooks.yml.
Subsequent creation requests will use the configuration stored or sensible defaults, if you want to overwrite these you can update the notebooks.yml file or pass in the --configure flag.
More information about the CLI command is available here.
In the UI, you can go to the notebook tab and click create a new notebook to start a plain notebook.
This will launch a popup form in order to configure your notebook.
After clicking on create, we will launch a container containing an empty JupyterLab installation, without any project specific code.
You can define a default IAM identity to use within the scope of a project. In case your project has defined its default identity, it will be filled out automatically after choosing the project and environment for your notebook.
If you want to learn how you can set the default IAM identity, you can refer to the CLI documentation.
RBAC support
A notebook is associated with a user, a project and an environment. The user should be at least a contributor on both the project and the environment in order to create/interact with the notebook.
Administrators are able to see all the notebooks and can delete them, but they are not allowed to access JupyterLab itself. This means they cannot change the notebook or source files of other users.
Data exploration flow
1. Using JupyterLab, you can experiment with your data interactively with notebook files or adjust the source code of your project.
2. If you have made relevant changes, you can download the current state to your local filesystem by running:
conveyor notebook download --env ENV --name NAME
This will only download files in the src and notebooks folder, as specified when configuring your notebook.
3. When you will not be working with the notebook for some time, but do want to keep it's current state, you can stop it:
conveyor notebook stop --env ENV --name NAME
This will save the project state as well as the virtual environment. If you want to continue from where you left, you should execute the following command:
conveyor notebook start --env ENV --name NAME
4. When you are finished working with your notebook, you can delete it:
conveyor notebook delete --env ENV --name NAME
To keep costs under control when working with notebooks, we added a maxIdleTime property.
This indicates how long an inactive notebooks should remain running.
After this period, they will be stopped automatically, it will however save your current working state.
Conveyor notebook UI
In Conveyor, we added some additional pages to interact with notebooks. We created a new tab, where you can see your running notebooks. Admin users are able to see and delete all notebooks but are not allowed to access JupyterLab itself.
The overview page contains a link to JupyterLab and a button to delete the notebook. Additionally, you can also click on the notebook for more detailed information as well as an embedded version of JupyterLab.
