DEV Community

Cover image for How to Use Apache Airflow
Ahmet Can Aydemir
Ahmet Can Aydemir

Posted on • Edited on • Originally published at ahmetcan.org

How to Use Apache Airflow

#airflow #apache #orchestration #etl

Airflow is an orchestration tool that ensures that tasks are running at the right time, in the correct order, and in the right way.

Why You Need Apache Airflow

ETL

Imagine you have a data pipeline like the one above.

  • What if an error occurs in any of these stages? There may be an error in the API from which you are pulling the data, there may be an error while processing the data, or there may be an error while saving to the DB.
  • If you have a lot of data pipelines like this, it will eventually become overwhelming.

Roughly, as in the example above, taking the data from a source and saving it to the target after certain operations are called ETL (Extract Transform Load). Such transactions can be managed in an advanced way by using Airflow.

Benefits of Apache Airflow

Dynamic: What can be done with Python also can be done with Airflow. As a result, Airflow provides tremendous dynamics when creating our tasks.

Scalable: As many tasks as desired can be easily run in parallel.

User Interface: Airflow has a useful UI. Errors that occur in data pipelines and where they occur can be easily observed. Problematic tasks can be restarted etc.

Extensible: No need to wait for Airflow update when a new tool comes out. You can write your plugins and integrate them easily.

Apache Airflow’s Core Components

Airflow's Core Components

Web Server: A Flask server that serves the UI.

Scheduler: The daemon that schedules the workflows. It is the most important component of Airflow.

Metastore: Database where metadata is stored.

Executor: Class that defines how tasks should work.

Worker: The process or sub-process executing the task.

DAG (Directed Acyclic Graph)

A DAG (Directed Acyclic Graph) is the core concept of Airflow, collecting Tasks together, organized with dependencies and relationships to say how they should run.

DAG

  • Example DAG above defines four Tasks - A, B, C, and D - and dictates the order in which they have to run, and which tasks depend on what others. It will also say how often to run the DAG - maybe “every 5 minutes starting tomorrow”, or “every day since January 1st, 2020”.
  • The DAG itself doesn’t care about what is happening inside the tasks; it is merely concerned with how to execute them - the order to run them in, how many times to retry them, if they have timeouts, and so on.

Basic DAG example:

with DAG("my_dag_name",
        start_date=datetime(2020, 1, 1),
    schedule_interval="@daily",
        catchup=False)as dag:
    op = DummyOperator(task_id="task")
Enter fullscreen mode Exit fullscreen mode

Operators

Operators are wrappers that cover the task. By default, there are many different types of operators and can be viewed at this link. In addition, much more can be added as needed.

  1. Action Operators: Operators which executes functions or commands eg. bash scripts, python code
  2. Transfer Operators: Operators for moving data from source to destination eg. ElasticSearch to Mysql
  3. Sensor Operators: Operators which wait for something to happen before moving on to another task, eg. checking the file in the directory and continuing to the other task after that.

What Isn’t Airflow?

Airflow is not a data streaming solution or data processing framework. If you need to process data every second, instead of using Airflow, Spark or Flink would be a better solution. If terabytes of data are being processed, it is recommended to run the Spark job with the operator in Airflow.

Apache Airflow Setup

Although Airflow can be installed with Docker, Kubernetes or different methods, in this article, we will install it locally.

# Airflow needs a home. `~/airflow` is the default, but you can put it
# somewhere else if you prefer (optional)
export AIRFLOW_HOME=~/airflow

# Install Airflow using the constraints file. For this example, we will use Python 3.6 and Airflow 2.2.4
pip install "apache-airflow==${AIRFLOW_VERSION}" --constraint "https://raw.githubusercontent.com/apache/airflow/constraints-2.2.4/constraints-3.6.txt"

# The Standalone command will initialize the database, make a user,
# and start all components for you.
airflow standalone

# Visit localhost:8080 in the browser and use the admin account details
# shown on the terminal to login.
# Enable the example_bash_operator dag on the home page
Enter fullscreen mode Exit fullscreen mode

First Pipeline with Apache Airflow

Pipeline

For example, let's create a data pipeline like the one in the figure using Airflow. For this, let's create a new python file in ~/airflow/dags directory eg: user_processing.py

1. Importing Modules 


from datetime import datetime
import json
from pandas import json_normalize

# DAG object; We will need this to create DAG
from airflow import DAG

# Operators
from airflow.operators.bash import BashOperator
from airflow.operators.python import PythonOperator
from airflow.providers.sqlite.operators.sqlite import SqliteOperator
from airflow.providers.http.operators.http import SimpleHttpOperator

# Sensors
from airflow.providers.http.sensors.http import HttpSensor
Enter fullscreen mode Exit fullscreen mode

2. Instantiate a DAG

We’ll need a DAG object to nest our tasks into. Here we pass a string that defines the dag_id, which serves as a unique identifier for your DAG. We also pass a default argument dictionary and define a schedule_interval of 1 day for the DAG.

default_args = {
    'start_date': datetime(2020, 1, 1)
}

with DAG('user_processing', schedule_interval='@daily',
        default_args=default_args,
        catchup=False) as dag:

        # Define tasks/operators
Enter fullscreen mode Exit fullscreen mode

3. Creating the SQLite Table

We can use SqliteOperator for this task. But SqliteOperator will need a parameter named conn_id, so first let's create a connection for SQLite DB.

In Airflow, connections are kept in the metadata database. We will need to do the same for the HTTPOperator.

We can use UI to create a connection

  1. Go to http://localhost:8080/connection/add (Admin / Connections / +)
  2. Type Conn Id field a unique name e.g.: db_sqlite
  3. Select Conn Type field as SQLite
  4. In the Host field, type the path to the SQLite database file. In this example, we can use Airflow's metadata database instead of creating a new SQLite database file. /home/airflow/airflow.db
  5. Save

After this process, we can create the operator. Each operator must have a unique task_id. SqliteOperator also has sqlite_conn_id and sql parameters. We will use the db_sqlite as the sqlite_conn_id (see step 2).

In SQL code; if the users table does not already exist; It creates a users table with fields firstname, lastname, country, username, password and email.

 creating_table = SqliteOperator(
        task_id='creating_table',
        sqlite_conn_id='db_sqlite',
        sql='''
            CREATE TABLE IF NOT EXISTS users (
                firstname TEXT NOT NULL,
                lastname TEXT NOT NULL,
                country TEXT NOT NULL,
                username TEXT NOT NULL,
                password TEXT NOT NULL,
                email TEXT NOT NULL PRIMARY KEY
            );
        ''' 
    )
Enter fullscreen mode Exit fullscreen mode

4. Verify API

We can add a sensor to check if there is a problem with the API by making an HTTP request. HttpSensor will be useful for this, but HttpSensor needs http_conn_id parameter. So we have to create a new connection as we did in the previous step. Unlike before, this time we will create a connection with HTTP type, not an SQLite type.

We can use UI to create a connection

  1. Go to http://localhost:8080/connection/add (Admin / Connections / +)
  2. Type Conn Id field a unique name e.g.: user_api
  3. Select Conn Type field as HTTP
  4. In the Host field, enter the address of the API that brings random users. eg. https://randomuser.me/
  5. Save

Now we can create the sensor. Since the API we use works at [https://randomuser.me/api/](https://randomuser.me/api), we used api/ in the endpoint parameter.

is_api_available = HttpSensor(
        task_id='is_api_available',
        http_conn_id='user_api',
        endpoint='api/'
    )
Enter fullscreen mode Exit fullscreen mode

5. Retrieve Random Users from API

With using SimpleHttpOperator we send a GET HTTP request to the user_api connection we created before. With response_filter we convert the JSON string in the response.text field into a python object.

 extracting_user = SimpleHttpOperator(
        task_id='extracting_user',
        http_conn_id='user_api',
        endpoint='api/',
        method='GET',
        response_filter=lambda response: json.loads(response.text),
        log_response=True
    )
Enter fullscreen mode Exit fullscreen mode

6. Processing the API Result

With using PythonOperator, we will keep only the desired fields of the python object that we obtained in step 5. And we will save the result as csv file.

Any python function can be run directly with the python_callable parameter.

The important part here is how we will get the response from the previous section. For this, we can send the task instance (ti) object as a parameter to the python function and get the results of the task_ids with users = ti.xcom_pull(task_ids=['extracting_user']).

processing_user = PythonOperator(
        task_id='processing_user',
        python_callable=_processing_user
    )

# Somewhere outside the scope of DAG
def _processing_user(ti):
    users = ti.xcom_pull(task_ids=['extracting_user'])
    if not len(users) or 'results' not in users[0]:
        raise ValueError('User is empty')
    user = users[0]['results'][0]
    processed_user = json_normalize({
        'firstname': user['name']['first'],
        'lastname': user['name']['last'],
        'country': user['location']['country'],
        'username': user['login']['username'],
        'password': user['login']['password'],
        'email': user['email']
    })
    processed_user.to_csv('/tmp/processed_user.csv', index=None, header=False)
Enter fullscreen mode Exit fullscreen mode

7. Saving Processed Users to SQLite DB

We can use BashOperator for trying new operators. BashOperator allows us to run bash commands.

storing_user = BashOperator(
        task_id='storing_user',
        bash_command='echo -e ".separator ","\n.import /tmp/processed_user.csv users" | sqlite3 /home/airflow/airflow/airflow.db'
    )
Enter fullscreen mode Exit fullscreen mode

8. Relationship Between Tasks

Simply, the direction of the tasks and their connection with each other can be defined as follows.

# Somewhere outside the scope of DAG
creating_table >> is_api_available >> extracting_user >> processing_user >> storing_user
Enter fullscreen mode Exit fullscreen mode

Final DAG File 

from datetime import datetime
import json
from pandas import json_normalize

from airflow import DAG

from airflow.operators.bash import BashOperator
from airflow.operators.python import PythonOperator
from airflow.providers.sqlite.operators.sqlite import SqliteOperator
from airflow.providers.http.operators.http import SimpleHttpOperator
from airflow.providers.http.sensors.http import HttpSensor

default_args = {
    'start_date': datetime(2020, 1, 1)
}

def _processing_user(ti):
    users = ti.xcom_pull(task_ids=['extracting_user'])
    if not len(users) or 'results' not in users[0]:
        raise ValueError('User is empty')
    user = users[0]['results'][0]
    processed_user = json_normalize({
        'firstname': user['name']['first'],
        'lastname': user['name']['last'],
        'country': user['location']['country'],
        'username': user['login']['username'],
        'password': user['login']['password'],
        'email': user['email']
    })
    processed_user.to_csv('/tmp/processed_user.csv', index=None, header=False)


with DAG('user_processing', schedule_interval='@daily',
        default_args=default_args,
        catchup=False) as dag:
        # Define tasks/operators

    creating_table = SqliteOperator(
        task_id='creating_table',
        sqlite_conn_id='db_sqlite',
        sql='''
            CREATE TABLE IF NOT EXISTS users (
                firstname TEXT NOT NULL,
                lastname TEXT NOT NULL,
                country TEXT NOT NULL,
                username TEXT NOT NULL,
                password TEXT NOT NULL,
                email TEXT NOT NULL PRIMARY KEY
            );
        ''' 
    )

    is_api_available = HttpSensor(
        task_id='is_api_available',
        http_conn_id='user_api',
        endpoint='api/'
    )

    extracting_user = SimpleHttpOperator(
        task_id='extracting_user',
        http_conn_id='user_api',
        endpoint='api/',
        method='GET',
        response_filter=lambda response: json.loads(response.text),
        log_response=True
    )

    processing_user = PythonOperator(
        task_id='processing_user',
        python_callable=_processing_user
    )

    storing_user = BashOperator(
        task_id='storing_user',
        bash_command='echo -e ".separator ","\n.import /tmp/processed_user.csv users" | sqlite3 /home/airflow/airflow/airflow.db'
    )

creating_table >> is_api_available >> extracting_user >> processing_user >> storing_user
Enter fullscreen mode Exit fullscreen mode

Conclusion

To understand the concept, we have defined a simple and parallelism-free data pipeline. Much more complex, parallel tasks also can be created using Airflow. Thank you for reading!

Screenshots

Show Screenshots
Graph View

Tree View

Gantt

Log

Homepage

References

Top comments (0)

Read next

r9n profile image

Flappy Bird Multiplyer

Ronaldo Modesto -

xun_wang_6384a403f9817c2 profile image

HarmonyOS Case Practice: Design and Implementation of a High-Concurrency Data Acquisition System

SameX -

2001asjad profile image

Creating and Managing Tokens with KALP Studio's Token Studio

Asjad Ahmed Khan -

alikolahdoozan profile image

Is .NET 9 beneficial for Blazor?

Ali Kolahdoozan -