def update_athena_partition(*args, **kwargs):
execution_date = datetime.datetime.strptime(kwargs['ds'], '%Y-%m-%d')
execution_month = execution_date.month
execution_year = execution_date.year
s3_prefix = Variable.get('bikeshare_s3_prefix')
logs_location = Variable.get('bikeshare_athena_logs')
athena_db = Variable.get('bikeshare_athena_db')
bucket_name = Variable.get('bikeshare_bucket_name')
athena_table_name = Variable.get('bikeshare_athena_table')
file_location = 's3://bikeshare-data-copy/' + s3_prefix + f'year={execution_year}/month={execution_month}/'
result_configuration = {"OutputLocation": logs_location}
partition_update_query = """
ALTER TABLE {} add partition (year="{}", month='{}')
location "{}";
"""
athena_hook = AWSAthenaHook(aws_conn_id='aws_credentials')
athena_hook.run_query(partition_update_query.format(
athena_table_name, execution_year, execution_month, file_location),
result_configuration=result_configuration,
query_context={"Database": athena_db},
client_request_token=str(uuid4()))
def check_data_in_redshift(*args, **kwargs):
execution_date = datetime.datetime.strptime(kwargs['ds'], '%Y-%m-%d')
execution_month = execution_date.month
execution_year = execution_date.year
bucket_name = Variable.get("bikeshare_bucket_name")
logs_location = Variable.get("bikeshare_athena_logs")
athena_db = Variable.get("bikeshare_athena_db")
num_records_athena_query = f"""
SELECT COUNT(*) FROM trips
WHERE year = {execution_year} AND month = {execution_month}
"""
num_records_redshift_query = f"""
SELECT COUNT(*) FROM trips
WHERE date_part(year, trips.start_time) = {execution_year} AND date_part(month, trips.start_time) = {execution_month}
"""
athena_hook = AWSAthenaHook(aws_conn_id='aws_credentials')
query_id = athena_hook.run_query(
query=num_records_athena_query,
query_context={"Database": athena_db},
result_configuration={"OutputLocation": logs_location},
client_request_token=str(uuid4()))
time.sleep(20)
athena_query_results = athena_hook.get_query_results(
query_execution_id=query_id)
num_records_in_s3 = int(athena_query_results['ResultSet']['Rows'][1]
['Data'][0]['VarCharValue'])
redshift_hook = PostgresHook('redshift_connection')
redshift_query_results = redshift_hook.get_records(
sql=num_records_redshift_query)
num_records_in_redshift = int(redshift_query_results[0][0])
logging.info(f"Number of records in S3 - {num_records_in_s3}")
logging.info(f"Number of records in Redshift - {num_records_in_redshift}")
if num_records_in_s3 == num_records_in_redshift:
logging.info(
f"Successfully Copied data for the year - {execution_year}, month - {execution_month}"
)
else:
raise Partial_Data_Missing
def get_hook(self):
return AWSAthenaHook(self.aws_conn_id, self.sleep_time)
