Skip to main content
Version: 0.8.0

Using Redpanda to test FastKafka

What is FastKafka?​

FastKafka is a powerful and easy-to-use Python library for building asynchronous services that interact with Kafka topics. Built on top of Pydantic, AIOKafka and AsyncAPI, FastKafka simplifies the process of writing producers and consumers for Kafka topics, handling all the parsing, networking, task scheduling and data generation automatically. With FastKafka, you can quickly prototype and develop high-performance Kafka-based services with minimal code, making it an ideal choice for developers looking to streamline their workflow and accelerate their projects.

What is Redpanda?​

Redpanda is a drop-in replacement for Kafka. Most of the Kafka tools work out of the box with Redpanda.

From redpanda.com:

Redpanda is a Kafka®-compatible streaming data platform that is proven to be 10x faster and 6x lower in total costs. It is also JVM-free, ZooKeeper®-free, Jepsen-tested and source available.

Some of the advantages of Redpanda over Kafka are

  1. A single binary with built-in everything, no ZooKeeper® or JVM needed.
  2. Costs upto 6X less than Kafka.
  3. Up to 10x lower average latencies and up to 6x faster Kafka transactions without compromising correctness.

To learn more about Redpanda, please visit their website or checkout this blog post comparing Redpanda and Kafka’s performance benchmarks.

Example repo​

A sample FastKafka-based library that uses Redpanda for testing, based on this guide, can be found here.

The process​

Here are the steps we’ll be walking through to build our example:

  1. Set up the prerequisites.
  2. Clone the example repo.
  3. Explain how to write an application using FastKafka.
  4. Explain how to write a test case to test FastKafka with Redpanda.
  5. Run the test case and produce/consume messages.

1. Prerequisites​

Before starting, make sure you have the following prerequisites set up:

  1. Python 3.x: A Python 3.x installation is required to run FastKafka. You can download the latest version of Python from the official website. You’ll also need to have pip installed and updated, which is Python’s package installer.
  2. Docker Desktop: Docker is used to run Redpanda, which is required for testing FastKafka. You can download and install Docker Desktop from the official website.
  3. Git: You’ll need to have Git installed to clone the example repo. You can download Git from the official website.

2. Cloning and setting up the example repo​

To get started with the example code, clone the GitHub repository by running the following command in your terminal:

git clone https://github.com/airtai/sample_fastkafka_with_redpanda.git
cd sample_fastkafka_with_redpanda

This will create a new directory called sample_fastkafka_with_redpanda and download all the necessary files.

Create a virtual environment​

Before writing any code, let’s create a new virtual environment for our project.

A virtual environment is an isolated environment for a Python project, which allows you to manage project-specific dependencies and avoid conflicts between different projects.

To create a new virtual environment, run the following commands in your terminal:

python3 -m venv venv

This will create a new directory called venv in your project directory, which will contain the virtual environment.

To activate the virtual environment, run the following command:

source venv/bin/activate

This will change your shell’s prompt to indicate that you are now working inside the virtual environment.

Finally, run the following command to upgrade pip, the Python package installer:

pip install --upgrade pip

Install Python dependencies​

Next, let’s install the required Python dependencies. In this guide, we’ll be using FastKafka to write our application code and pytest and pytest-asyncio to test it.

You can install the dependencies from the requirements.txt file provided in the cloned repository by running:

pip install -r requirements.txt

This will install all the required packages and their dependencies.

3. Writing server code​

The application.py file in the cloned repository demonstrates how to use FastKafka to consume messages from a Kafka topic, make predictions using a predictive model, and publish the predictions to another Kafka topic. Here is an explanation of the code:

Preparing the demo model​

First we will prepare our model using the Iris dataset so that we can demonstrate the predictions using FastKafka. The following call downloads the dataset and trains the model.

We will wrap the model creation into a lifespan of our app so that the model is created just before the app is started.

from contextlib import asynccontextmanager

from sklearn.datasets import load_iris
from sklearn.linear_model import LogisticRegression

from fastkafka import FastKafka

ml_models = {}


@asynccontextmanager
async def lifespan(app: FastKafka):
    # Load the ML model
    X, y = load_iris(return_X_y=True)
    ml_models["iris_predictor"] = LogisticRegression(random_state=0, max_iter=500).fit(
        X, y
    )
    yield
    # Clean up the ML models and release the resources
    ml_models.clear()

Messages​

FastKafka uses Pydantic to parse input JSON-encoded data into Python objects, making it easy to work with structured data in your Kafka-based applications. Pydantic’s BaseModel class allows you to define messages using a declarative syntax, making it easy to specify the fields and types of your messages.

This example defines two message classes for use in a FastKafka application:

  • The IrisInputData class is used to represent input data for a predictive model. It has four fields of type NonNegativeFloat, which is a subclass of float that only allows non-negative floating point values.

  • The IrisPrediction class is used to represent the output of the predictive model. It has a single field species of type string representing the predicted species.

These message classes will be used to parse and validate incoming data in Kafka consumers and producers.

from pydantic import BaseModel, Field, NonNegativeFloat


class IrisInputData(BaseModel):
    sepal_length: NonNegativeFloat = Field(
        ..., example=0.5, description="Sepal length in cm"
    )
    sepal_width: NonNegativeFloat = Field(
        ..., example=0.5, description="Sepal width in cm"
    )
    petal_length: NonNegativeFloat = Field(
        ..., example=0.5, description="Petal length in cm"
    )
    petal_width: NonNegativeFloat = Field(
        ..., example=0.5, description="Petal width in cm"
    )


class IrisPrediction(BaseModel):
    species: str = Field(..., example="setosa", description="Predicted species")

Application​

This example shows how to initialize a FastKafka application.

It starts by defining a dictionary called kafka_brokers, which contains two entries: "localhost" and "production", specifying local development and production Kafka brokers. Each entry specifies the URL, port, and other details of a Kafka broker. This dictionary is used both to generate documentation and to later run the server against one of the given kafka broker.

Next, an instance of the FastKafka class is initialized with the minimum required arguments:

  • kafka_brokers: a dictionary used for generating documentation
from fastkafka import FastKafka

kafka_brokers = {
    "localhost": {
        "url": "localhost",
        "description": "local development kafka broker",
        "port": 9092,
    },
    "production": {
        "url": "kafka.airt.ai",
        "description": "production kafka broker",
        "port": 9092,
        "protocol": "kafka-secure",
        "security": {"type": "plain"},
    },
}

kafka_app = FastKafka(
    title="Iris predictions",
    kafka_brokers=kafka_brokers,
    lifespan=lifespan,
)

Function decorators​

FastKafka provides convenient function decorators @kafka_app.consumes and @kafka_app.produces to allow you to delegate the actual process of

  • consuming and producing data to Kafka, and

  • decoding and encoding JSON encode messages

from user defined functions to the framework. The FastKafka framework delegates these jobs to AIOKafka and Pydantic libraries.

These decorators make it easy to specify the processing logic for your Kafka consumers and producers, allowing you to focus on the core business logic of your application without worrying about the underlying Kafka integration.

This following example shows how to use the @kafka_app.consumes and @kafka_app.produces decorators in a FastKafka application:

  • The @kafka_app.consumes decorator is applied to the on_input_data function, which specifies that this function should be called whenever a message is received on the “input_data" Kafka topic. The on_input_data function takes a single argument which is expected to be an instance of the IrisInputData message class. Specifying the type of the single argument is instructing the Pydantic to use IrisInputData.parse_raw() on the consumed message before passing it to the user defined function on_input_data.

  • The @produces decorator is applied to the to_predictions function, which specifies that this function should produce a message to the “predictions" Kafka topic whenever it is called. The to_predictions function takes a single integer argument species_class representing one of three possible strign values predicted by the mdoel. It creates a new IrisPrediction message using this value and then returns it. The framework will call the IrisPrediction.json().encode("utf-8") function on the returned value and produce it to the specified topic.

@kafka_app.consumes(topic="input_data", auto_offset_reset="latest")
async def on_input_data(msg: IrisInputData):
    species_class = ml_models["iris_predictor"].predict(
        [[msg.sepal_length, msg.sepal_width, msg.petal_length, msg.petal_width]]
    )[0]

    await to_predictions(species_class)


@kafka_app.produces(topic="predictions")
async def to_predictions(species_class: int) -> IrisPrediction:
    iris_species = ["setosa", "versicolor", "virginica"]

    prediction = IrisPrediction(species=iris_species[species_class])
    return prediction

4. Writing the test code​

The service can be tested using the Tester instance which can be configured to start a Redpanda broker for testing purposes. The test.py file in the cloned repository contains the following code for testing.

import pytest
from application import IrisInputData, IrisPrediction, kafka_app

from fastkafka.testing import Tester

msg = IrisInputData(
    sepal_length=0.1,
    sepal_width=0.2,
    petal_length=0.3,
    petal_width=0.4,
)


@pytest.mark.asyncio
async def test():
    # Start Tester app and create local Redpanda broker for testing
    async with Tester(kafka_app).using_local_redpanda(
        tag="v23.1.2", listener_port=9092
    ) as tester:
        # Send IrisInputData message to input_data topic
        await tester.to_input_data(msg)

        # Assert that the kafka_app responded with IrisPrediction in predictions topic
        await tester.awaited_mocks.on_predictions.assert_awaited_with(
            IrisPrediction(species="setosa"), timeout=2
        )

The Tester module utilizes uses LocalRedpandaBroker to start and stop a Redpanda broker for testing purposes using Docker

5. Running the tests​

We can run the tests which is in test.py file by executing the following command:

pytest test.py

This will start a Redpanda broker using Docker and executes tests. The output of the command is:

(venv) fastkafka@airt-ai:~/dev/sample_fastkafka_with_redpanda$ pytest
============================== test session starts ===============================
platform linux -- Python 3.10.6, pytest-7.2.2, pluggy-1.0.0
rootdir: /home/kumaran/dev/sample_fastkafka_with_redpanda, configfile: pytest.ini, testpaths: test.py
plugins: asyncio-0.21.0, anyio-3.6.2
asyncio: mode=strict
collected 1 item                                                                 

test.py .                                                                  [100%]

=============================== 1 passed in 7.28s ================================
(venv) fastkafka@airt-ai:~/dev/sample_fastkafka_with_redpanda$

Running the tests with the Redpanda broker ensures that your code is working correctly with a real Kafka-like message broker, making your tests more reliable.

Recap​

We have created an Iris classification model and encapulated it into our FastKafka application. The app will consume the IrisInputData from the input_data topic and produce the predictions to predictions topic.

To test the app we have:

  1. Created the app

  2. Started our Tester class with Redpanda broker which mirrors the developed app topics for testing purposes

  3. Sent IrisInputData message to input_data topic

  4. Asserted and checked that the developed iris classification service has reacted to IrisInputData message