How do we know that an Airflow DAG ran successfully? For sure, we should check the DAG run status and the statuses of all tasks, but is it enough to see that were no errors? What if everything is ok only because there was no data in the data source?

This article shows how to use the AWSAthenaOperator to check that an Athena SQL statement returns data. I use it to verify that we have input data in a given partition before running a DAG and ensure that data analysts can access the output.

First, we need to import the BaseOperator and the AWSAthenaOperator:

from airflow.contrib.operators.aws_athena_operator import AWSAthenaOperator
from airflow.models import BaseOperator

Now, we will create a custom Airflow operator:

class AWSAthenaTableNotEmpty(BaseOperator):
    template_fields = ('query', 'database')
    template_ext = ('.sql')

    def __init__(
            query: str,
            database: str,
            s3_output_locations: str,

        super().__init__(*args, **kwargs)
        self.extra_dag_args = args
        self.extra_dag_kwargs = kwargs

        self.query = query
        self.database = database
        self.output_location = "s3://" + s3_output_location

        self.run_query_task = None

In the execute function, we will run the query, retrieve the results, and check that we received at least one row.

def execute(self, context):
    self.run_query_task = AWSAthenaOperator(query=self.query,

    query_execution_id = self.run_query_task.query_execution_id
    athena_hook = self.run_query_task.get_hook()
    raw_results = athena_hook.get_query_results(query_execution_id)

    result_count = len(raw_results['ResultSet']['Rows'][1:])

    assert result_count > 0,\
        'There are no rows in the table! ' \
        f'Database: {self.database} ' \
        f'Query: {self.query} '

We should also make sure that Airflow can stop the task, so we must implement the on_kill function:

def on_kill(self):
    if self.run_query_task:

How to use the AWSAthenaTableNotEmpty operator

To check that we have the input data for an Airflow DAG or that it has produced some results, we should create an instance of the operator and put it either at the beginning or the end of the DAG:

verify_input = AWSAthenaTableNotEmpty(
        query=f'select * from some_input_table '
                  "WHERE year = '{{ macros.ds_format(ds, '%Y-%m-%d', '%Y') }}' "
                  "AND month = '{{ macros.ds_format(ds, '%Y-%m-%d', '%m') }}' "
                  "AND day = '{{ macros.ds_format(ds, '%Y-%m-%d', '%d') }}' "
                  "LIMIT 1",
        database='{{ params.database_name }}')

Older post

How to start an AWS Glue Crawler to refresh Athena tables using boto3

How to create and start an AWS Glue Crawler from Python code using boto3

Newer post

How to configure Spark to maximize resource usage while using AWS EMR

How to configure EMR to use all available resources when running a Spark cluster