Submit a spark job from Airflow to external spark container

Question

I have a spark and airflow cluster which is built with docker swarm. Airflow container cannot contain spark-submit as I expect.

I am using following images which exist in github

Spark: big-data-europe/docker-hadoop-spark-workbench

Airflow: puckel/docker-airflow (CeleryExecutor)

I prepared a .py file and add it under dags folder.

from airflow import DAG

from airflow.contrib.operators.spark_submit_operator import SparkSubmitOperator
from datetime import datetime, timedelta


args = {'owner': 'airflow', 'start_date': datetime(2018, 9, 24) }

dag = DAG('spark_example_new', default_args=args, schedule_interval="@once")

operator = SparkSubmitOperator(task_id='spark_submit_job', conn_id='spark_default', java_class='Main', application='/SimpleSpark.jar', name='airflow-spark-example',conf={'master':'spark://master:7077'},
        dag=dag)

I also configure the connection as folows in web site:

Master is the hostname of spark master container.

But it does not find the spark-submit, it produces following error:

[2018-09-24 08:48:14,063] {{logging_mixin.py:95}} INFO - [2018-09-24 08:48:14,062] {{spark_submit_hook.py:283}} INFO - Spark-Submit cmd: ['spark-submit', '--master', 'spark://master:7077', '--conf', 'master=spark://master:7077', '--name', 'airflow-spark-example', '--class', 'Main', '--queue', 'root.default', '/SimpleSpark.jar']

[2018-09-24 08:48:14,067] {{models.py:1736}} ERROR - [Errno 2] No such file or directory: 'spark-submit': 'spark-submit'
Traceback (most recent call last):
  File "/usr/local/lib/python3.6/site-packages/airflow/models.py", line 1633, in _run_raw_task
    result = task_copy.execute(context=context)
  File "/usr/local/lib/python3.6/site-packages/airflow/contrib/operators/spark_submit_operator.py", line 168, in execute
    self._hook.submit(self._application)
  File "/usr/local/lib/python3.6/site-packages/airflow/contrib/hooks/spark_submit_hook.py", line 330, in submit
    **kwargs)
  File "/usr/local/lib/python3.6/subprocess.py", line 709, in __init__
    restore_signals, start_new_session)
  File "/usr/local/lib/python3.6/subprocess.py", line 1344, in _execute_child
    raise child_exception_type(errno_num, err_msg, err_filename)
FileNotFoundError: [Errno 2] No such file or directory: 'spark-submit': 'spark-submit'

Answer 1

As far as I know puckel/docker-airflow uses Python slim image(https://hub.docker.com/_/python/). This image does not contain the common packages and only contains the minimal packages needed to run python. Hence, you will need to extend the image and install spark-submit on your container.

Edit: Airflow does need spark binaries in the container to run SparkSubmitOperator as documented here.

The other approach you can use is to use SSHOperator to run spark-submit command on an external VM by SSHing into a remote machine. But here as well SSH should be available which isn't available in Puckel Airflow.

Answer 2

• This is late answer

• you should install apache-airflow-providers-apache-spark

• So you should create a file called 'requirements.txt'

• add apache-airflow-providers-apache-spark in the file requirements.txt

• Create a Dockerfile like this

   FROM apache/airflow:2.2.3
  
   # Install OpenJDK-11
   RUN apt update && \
       apt-get install -y openjdk-11-jdk && \
       apt-get install -y ant && \
       apt-get clean;
  
   # Set JAVA_HOME
   ENV JAVA_HOME /usr/lib/jvm/java-11-openjdk-amd64/
   RUN export JAVA_HOME
  
   USER airflow
  
   COPY requirements.txt .
   RUN pip install -r requirements.txt
  

• in the docker-compose.yml comment the line :

  # image: ${AIRFLOW_IMAGE_NAME:-apache/airflow:2.2.3}

• and uncomment the line build .

Finally run

docker-compose build
docker-compose up