luigi.contrib.spark module

class luigi.contrib.spark.SparkSubmitTask(*args, **kwargs)

Bases: luigi.contrib.external_program.ExternalProgramTask

Template task for running a Spark job

Supports running jobs on Spark local, standalone, Mesos or Yarn

See http://spark.apache.org/docs/latest/submitting-applications.html for more information

name = None

entry_class = None

app = None

always_log_stderr = False

stream_for_searching_tracking_url = 'stderr'

tracking_url_pattern

app_options()

Subclass this method to map your task parameters to the app’s arguments

pyspark_python

pyspark_driver_python

hadoop_user_name

spark_version

spark_submit

master

deploy_mode

jars

packages

py_files

files

conf

properties_file

driver_memory

driver_java_options

driver_library_path

driver_class_path

executor_memory

driver_cores

supervise

total_executor_cores

executor_cores

queue

num_executors

archives

hadoop_conf_dir

get_environment()

program_environment()

Override this method to control environment variables for the program

Returns:dict mapping environment variable names to values

program_args()

Override this method to map your task parameters to the program arguments

Returns:list to pass as args to subprocess.Popen

spark_command()

app_command()

class luigi.contrib.spark.PySparkTask(*args, **kwargs)

Bases: luigi.contrib.spark.SparkSubmitTask

Template task for running an inline PySpark job

Simply implement the main method in your subclass

You can optionally define package names to be distributed to the cluster with py_packages (uses luigi’s global py-packages configuration by default)

app = '/home/docs/checkouts/readthedocs.org/user_builds/luigi/envs/latest/lib/python2.7/site-packages/luigi-2.8.12-py2.7.egg/luigi/contrib/pyspark_runner.py'

name

py_packages

files

setup(conf)

Called by the pyspark_runner with a SparkConf instance that will be used to instantiate the SparkContext

Parameters:conf – SparkConf

setup_remote(sc)

main(sc, *args)

Called by the pyspark_runner with a SparkContext and any arguments returned by app_options()

Parameters:

• sc – SparkContext
• args – arguments list

app_command()

run()

The task run method, to be overridden in a subclass.

See Task.run