Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

docs: [FC-0074] explain how the event bus works and how to use it #426

Merged
merged 12 commits into from
Dec 11, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions docs/Makefile
Original file line number Diff line number Diff line change
Expand Up @@ -18,3 +18,6 @@ help:
# "make mode" option. $(O) is meant as a shortcut for $(SPHINXOPTS).
%: Makefile
@$(SPHINXBUILD) -M $@ "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O)

serve_docs: ## serve the built docs locally to preview the site in the browser
sphinx-autobuild . $(BUILDDIR)/html
Binary file added docs/_images/event-bus-overview.png
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
129 changes: 129 additions & 0 deletions docs/concepts/event-bus.rst
Original file line number Diff line number Diff line change
@@ -0,0 +1,129 @@
Open edX Event Bus
==================

Overview
--------

The suggested strategy for cross-service communication in the Open edX ecosystem is through an event-based architecture implemented via the :term:`Event Bus`. This functionality used for asynchronous communication between services is built on top of sending Open edX Events (Open edX-specific Django signals) within a service.

What is the Open edX Event Bus?
-------------------------------

The :term:`Event Bus` implements an event-driven architecture that enables asynchronous communication between services using the `publish/subscribe messaging pattern`_ (pub/sub). In the Open edX ecosystem, the event bus is used to broadcast Open edX Events to multiple services, allowing them to react to changes or actions in the system. The event bus is a key component of the Open edX architecture, enabling services to communicate without direct dependencies on each other.

.. image:: ../_images/event-bus-overview.png
:alt: Open edX Event Bus Overview
:align: center

Why use the Open edX Event Bus?
-------------------------------

The :term:`Event Bus` can help us achieve loose coupling between services, replacing blocking requests between services and large sync jobs, leading to a faster, more reliable, and more extensible system. See event messaging architectural goals highlighted in :doc:`openedx-proposals:architectural-decisions/oep-0041-arch-async-server-event-messaging` to read more about its benefits. Here's a brief summary of some key points:

* **Eliminate Blocking, Synchronous Requests**: reduce site outages when services become overloaded with requests by replacing synchronous calls with asynchronous communication.
* **Eliminate Expensive, Delayed, Batch Synchronization**: replace expensive batch processing with near real-time data updates.
* **Reduce the need for Plugins**: reduce the computational load for plugins that don't need to run in the same process by allowing cross-service communication of lifecycle events.

How Does the Open edX Event Bus Work?
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should we also include more detail on how the data is serialized/deserialized? Or is this a good high-level overview?

Either way, I'm writing some documents on adding event bus support to new events so that the serialization/deserialization details can be included there.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this is good for a high level overview. I think it would be nice to put something in here about the differences between the Event Bus and Celery tasks, and when to use each. It's one of the more common questions I get.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes! I'm going to add a comparison table or something similar about them. Thank you for the suggestion!

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I compared the event bus async communication style with async tasks here: fb51a60. Let me know if this is what you had in mind. Thanks!

-------------------------------------

The Open edX platform uses the ``OpenEdxPublicSignals`` (Open edX-specific Django Signals) to send events within a service. The event bus extends these signals, allowing them to be broadcasted and handled across multiple services. That's how Open edX Events are used for internal and external communication. For more details on how these Open edX-specific Django Signals are used by the event bus, refer to the :doc:`../decisions/0004-external-event-bus-and-django-signal-events` Architectural Decision Record (ADR).

Open edX Events provides an abstract implementation of the `publish/subscribe messaging pattern`_ (pub/sub) which is the chosen pattern for the event bus implementation as explained in :doc:`openedx-proposals:architectural-decisions/oep-0052-arch-event-bus-architecture`. It implements two abstract classes, `EventProducer`_ and `EventConsumer`_ which allow concrete implementations of the event bus based on different message brokers, such as Pulsar.

This abstraction allows for developers to implement their own concrete implementations of the event bus in their own plugins. There are currently two implementations of the event bus, Redis (`event-bus-redis`_) and Kafka (`event-bus-kafka`_). Redis streams is a data structure that acts like an append-only log, and Kafka is a distributed event streaming application. Both implementations handle event production and consumption with technology specific methods.

Architectural Diagram
*********************

These diagrams show what happens when an event is sent to the event bus. The event sending workflow follows the same steps as explained in :ref:`events-architecture`, with a key difference: when configured, the event bus recognizes events and publish them to the message broker for consumption by other services.

Components
~~~~~~~~~~

* **Service A (Producer)**: The service that emits the event. Developers may have emitted this event in a key section of the application logic, signaling that a specific action has occurred.
* **Service B (Consumer)**: The service that listens for the event and executes custom logic in response.
* **OpenEdxPublicSignal**: The class that implements all methods used to manage sending the event. This class inherits from Django's Signals class and adds Open edX-specific metadata and behaviors.
* **Event Producer**: The class in the :term:`Producer` service that sends events to the event bus. The producer serializes the event data and enriches it with relevant metadata for the consumer.
* **Event Consumer**: The class in the :term:`Consumer` service that receives events from the event bus. The consumer deserializes the :term:`message <Message>` and re-emits it as an event with the data that was transmitted.
* **Message Broker**: The :term:`message broker <Message Broker>` is responsible for storing and delivering messages between the producer and consumer.
* **Event Bus Plugin**: The concrete implementation of the event bus (EventProducer and EventConsumer) based on a specific :term:`message broker <Message Broker>`, such as Pulsar. The plugin handles event production and consumption with technology-specific methods.

Workflow
~~~~~~~~

.. image:: ../_images/event-bus-workflow-service-a.png
:alt: Open edX Event Bus Workflow (Service A)
:align: center

**From Service A (Producer)**

1. When the event is sent, a registered event receiver `general_signal_handler`_ is called to send the event to the event bus. This receiver is registered by the Django Signal mechanism when the ``openedx-events`` app is installed, and it listens for all Open edX Events.
2. The receiver checks the ``EVENT_BUS_PRODUCER_CONFIG`` to look for the ``event_type`` of the event that was sent.
3. If the event type is found and it's enabled for publishing in the configuration, the receiver obtains the configured producer class (``EventProducer``) from the concrete event bus implementation. For example, the producer class for Redis or Kafka implemented in their respective plugins.
4. The ``EventProducer`` serializes the event data and enriches it with relevant metadata, and then transforms it into a message that can be transmitted.
5. The producer uses its technology-specific ``send`` method to publish a message to the configured broker (e.g., Redis or Kafka).

.. image:: ../_images/event-bus-workflow-service-b.png
:alt: Open edX Event Bus Workflow (Service B)
:align: center

**From Service B (Consumer)**

1. A :term:`Worker` process in Service B runs indefinitely, checking the broker for new messages.
2. When a new message is found, the ``EventConsumer`` deserializes the message and re-emits it as an event with the data that was transmitted.
3. The event sending and processing workflow repeats in Service B.

This approach of producing events via settings with the generic handler was chosen to allow for flexibility in the event bus implementation. It allows developers to choose the event bus implementation that best fits their needs, and to easily switch between implementations if necessary. See more details in the :doc:`../decisions/0012-producing-to-event-bus-via-settings` Architectural Decision Record (ADR).

Event Bus vs Asynchronous Tasks
-------------------------------

Asynchronous tasks are used for long-running, resource-intensive operations that should not block the main thread of a service. The event bus is used for broadcasting messages to multiple services, allowing them to react to changes or actions in the system. Both can be used for asynchronous communication, but they serve different purposes and have different workflows.

In this diagram, you can see the difference between the two when it comes to asynchronous communication:

.. image:: ../_images/event-bus-vs-asynchronous-tasks.png
:alt: Open edX Event Bus vs Asynchronous Tasks
:align: center

In the upper part of the diagram, we have Service A and Service B. Service A is the producer of the event, and a :term:`Worker` of Service B is the consumer. This is the event bus workflow which allows asynchronous non-blocking communication between services:

- Service A sends the event as a message to the event bus and continues its execution, as we previously explained.
- Service B polls the message broker for new messages and converts them into ``OpenEdxPublicSignal``.
- Service B emits the event locally and registered Django Signal receivers can react to it.
- Service C can also listen to the event and react to it.

In the lower part of the diagram, we have the asynchronous tasks workflow:

- A worker of Service A picks up a task which communicates with Service B via a request-response mechanism, such as HTTP.
- The worker of Service A sends a request to Service B and waits for a response.
- Service B processes the request and sends a response back to the worker.
- The worker receives the response and continues with the next step in its processing.

If we were to introduce a Service C in this scenario, it would need to wait for the worker of Service A to finish processing the response from Service B and receive a response before it could continue.

This is an example of an asynchronous approach (from the producer point of view) to send messages to another services but with a blocking nature.

Use the Open edX Event bus instead of asynchronous tasks when:
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is great! A thing we don't talk about often, and I'm not sure we should, is that the event bus (especially Kafka) can be used to send event data out of the Open edX ecosystem entirely. Many different databases and products can consume Kafka events for various purposes, for instance Student Information Services or Contact Relationship Managers may listen to enrollment events to update their own state. Do you think it's worth a bullet point here about that?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

That's a pretty interesting use case we could include here: https://docs.openedx.org/projects/openedx-events/en/latest/reference/real-life-use-cases.html#cross-services-communication. I think it'd give community folks a more rounded view of all that can be done with the event bus, do you have more info on this use case? Although I could use the info in the comment.

For this document purpose, I added some high-level ideas of external services that can consume events sent through the event bus: ee3f8e5


- You want to send a message but don't need a response. For example, notifying other services of an event that occurred.
- You need to send a high volume of messages to a single or multiple services. For example, notifying when users visit a unit in a course or watch a video.
- You want to decouple services and avoid direct dependencies.
- You want to send events out of the Open edX ecosystem. For example, external databases or services that can consume events to update their own state.

When you need to send a message to a particular service and wait their response for further processing, use asynchronous tasks.

How is the Open edX Event Bus Used?
-----------------------------------

The event bus is used to broadcast Open edX Events to multiple services, allowing them to react to changes or actions in the system.

We encourage you to review the :doc:`../reference/real-life-use-cases` page for examples of how the event bus is used in the Open edX ecosystem by the community. Also, see the :doc:`../how-tos/using-the-event-bus` guide to get start sending events to the event bus.

.. _general_signal_handler: https://github.com/openedx/openedx-events/blob/main/openedx_events/apps.py#L16-L44
.. _EventProducer: https://github.com/openedx/openedx-events/blob/main/openedx_events/event_bus/__init__.py#L71-L91
.. _EventConsumer: https://github.com/openedx/openedx-events/blob/main/openedx_events/event_bus/__init__.py#L128-L139
.. _publish/subscribe messaging pattern: https://en.wikipedia.org/wiki/Publish%E2%80%93subscribe_pattern
.. _event-bus-redis: https://github.com/openedx/event-bus-redis/
.. _event-bus-kafka: https://github.com/openedx/event-bus-kafka/
1 change: 1 addition & 0 deletions docs/concepts/index.rst
Original file line number Diff line number Diff line change
Expand Up @@ -6,3 +6,4 @@ Concepts
:caption: Contents:

openedx-events
event-bus
7 changes: 4 additions & 3 deletions docs/concepts/openedx-events.rst
Original file line number Diff line number Diff line change
Expand Up @@ -15,13 +15,15 @@ These events are built on top of Django signals, inheriting their behavior while

Django includes a "signal dispatcher" which helps decoupled applications get notified when actions occur elsewhere in the framework. In a nutshell, signals allow certain senders to notify a set of receivers that some action has taken place. They're especially useful when many pieces of code may be interested in the same events.

Events are primarily used as a communication method between internal services by leveraging Django Signals and external services using the `Event Bus technology`_, making them the standard communication mechanism within the Open edX ecosystem.
Events are primarily used as a communication method between internal services by leveraging Django Signals and external services using the :doc:`../concepts/event-bus`, making them the standard communication mechanism within the Open edX ecosystem.

How do Open edX Events work?
----------------------------

Open edX Events are implemented by a class called `OpenEdxPublicSignal`_, which inherits from `Django's Signals class` and adds behaviors specific to the Open edX ecosystem. Thanks to this design, ``OpenEdxPublicSignal`` leverages the functionality of Django signals, allowing developers to apply their existing knowledge of the Django framework.

.. _events-architecture:

Architectural Diagram
*********************

Expand Down Expand Up @@ -71,14 +73,13 @@ The `Django Signals Documentation`_ provides a more detailed explanation of how
How are Open edX Events used?
-----------------------------

As mentioned previously, developers can listen to Open edX Events by registering signal receivers from their Open edX Django plugins that respond to the emitted events or by using the `Event Bus technology`_ to send events to external services.
As mentioned previously, developers can listen to Open edX Events by registering signal receivers from their Open edX Django plugins that respond to the emitted events or by using the :doc:`../concepts/event-bus` to send events to external services.

For more information on using Open edX Events, refer to the :doc:`../how-tos/using-events` how-to guide. We also encourage you to explore the :doc:`../reference/real-life-use-cases` section for real-life examples of how Open edX Events are used by the community.

.. _Django Signals Documentation: https://docs.djangoproject.com/en/4.2/topics/signals/
.. _triggering the COURSE_ENROLLMENT_CREATED event: https://github.com/openedx/edx-platform/blob/master/common/djangoapps/student/models/course_enrollment.py#L777-L795
.. _course_enrollment_post_save receiver: https://github.com/openedx/edx-platform/blob/master/openedx/core/djangoapps/notifications/handlers.py#L38-L53
.. _Event Bus technology: https://openedx.atlassian.net/wiki/spaces/AC/pages/3508699151/How+to+start+using+the+Event+Bus
.. _Django signals registry mechanism: https://docs.djangoproject.com/en/4.2/topics/signals/#listening-to-signals
.. _signal receivers in their plugins: https://edx.readthedocs.io/projects/edx-django-utils/en/latest/edx_django_utils.plugins.html#edx_django_utils.plugins.constants.PluginSignals
.. _Open edX Django plugins: https://edx.readthedocs.io/projects/edx-django-utils/en/latest/plugins/readme.html
Expand Down
7 changes: 7 additions & 0 deletions docs/conf.py
Original file line number Diff line number Diff line change
Expand Up @@ -38,6 +38,7 @@
'sphinxcontrib.mermaid',
'code_annotations.contrib.sphinx.extensions.openedx_events',
'sphinx.ext.intersphinx',
'code_annotations.contrib.sphinx.extensions.settings',
]

# Add any paths that contain templates here, relative to this directory.
Expand Down Expand Up @@ -131,4 +132,10 @@
f"https://docs.openedx.org/{rtd_language}/{rtd_version}",
None,
),
"openedx-proposals": (
# Not setting the version on purpose because we always want the latest version
# of OEPs
f"https://docs.openedx.org/projects/openedx-proposals/{rtd_language}/latest",
None,
),
}
42 changes: 0 additions & 42 deletions docs/how-tos/adding-events-to-event-bus.rst

This file was deleted.

2 changes: 1 addition & 1 deletion docs/how-tos/index.rst
Original file line number Diff line number Diff line change
Expand Up @@ -7,6 +7,6 @@ How-tos

creating-new-events
adding-events-to-a-service
adding-events-to-event-bus
using-events
using-the-event-bus
add-new-event-bus-concrete-implementation
Loading
Loading