Lifecycle management with KedroSession

Note

This documentation is based on Kedro 0.17.1. If you spot anything that is incorrect then please create an issue or pull request.

Overview

A KedroSession allows you to:

  • Manage the lifecycle of a Kedro run

  • Persist runtime parameters with corresponding session IDs

  • Traceback runtime parameters, such as CLI command flags and environment variables

KedroSession decouples Kedro’s library components, managed by KedroContext, and any session data (both static and dynamic data). As a result, Kedro components and plugins can access session data without the need to import the KedroContext object and library components.

The main methods and properties of KedroSession are:

  • create(): Create a new instance of KedroSession with session data

  • load_context(): Instantiate KedroContext object

  • close(): Close the current session — although we recommend that you use the session object as a context manager, which will call close() automatically, as opposed to calling the method explicitly

  • run(): Run the pipeline with the arguments provided; see Running pipelines for details

Create a session

The following code creates a KedroSession object as a context manager and runs a pipeline inside the context, with session data provided. The session automatically closes after exit:

from kedro.framework.session import KedroSession

with KedroSession.create("<your-kedro-project-package-name>") as session:
    session.run()

You need to tell KedroSession the package name of your Kedro project so it can load your settings, nodes and pipelines. Additionally, you can provide the following optional arguments in KedroSession.create():

  • project_path: Path to the project root directory

  • save_on_close: A boolean value to indicate whether or not to save the session to disk when it’s closed

  • env: Environment for the KedroContext

  • extra_params: Optional dictionary containing extra project parameters for the underlying KedroContext; if specified, this will update (and therefore take precedence over) parameters retrieved from the project configuration

When you want to access to the most recent session object, use the helper function get_current_session() as follows:

from kedro.framework.session import get_current_session

session = get_current_session()
context = session.load_context()
context.catalog.load("my_data").head()