spark-etl 0.0.10

Generic ETL Pipeline Framework for Apache Spark

Overview

Goal

spark_etl provide a platform independent way of building spark application.

Benefit

• Your application can be moved to different spark platform without change or with very little change.

Supported platform

• Local spark via pyspark package.
• Spark cluster with Livy Interface
• Oracle Dataflow

Concepts

Application

An application is the code for a spark job. It contains:

• A main.py file which contain the application entry
• A manifest.json file, which specify the metadata of the application.
• A requirements.txt file, which specify the application dependency.

See examples/myapp for example.

Build an application

To build an application, run

./etl.py -a build --app-dir <app-dir> --build-dir <build-dir>

• <app_dir> is the directory where your application is located.

• <build-dir> is the directory where you want your build to be deployed

  • Your build actually located at <build-dir>/<version>, where <version> is specified by application's manifest file

• Build is mostly platform independent. You need to depend on package oci-core if you intent to use oci Dataflow

Application entry signature

In your application's main.py, you shuold have a main function with the following signature:

• spark is the spark session object
• input_args a dict, is the argument user specified when running this job.
• sysops is the system options passed, it is platform specific.
• Your main function's return value will be returned from the job submitter to the caller.

def main(spark, input_args, sysops={}):
    # your code here

See here for example.

Deployer

• spark_etl support the following deployer
  • spark_etl.vendors.local.LocalDeployer
  • spark_etl.deployers.HDFSDeployer
  • spark_etl.vendors.oracle.DataflowDeployer

the etl.py command use the config file to decide which deployer to use

Job Submitter

• spark_etl support the following job submitter
  • spark_etl.vendors.local.PySparkJobSubmitter
  • spark_etl.job_submitters.livy_job_submitter.LivyJobSubmitter
  • spark_etl.vendors.oracle.DataflowJobSubmitter
• Job summiter's run function returns the retrun value from job's main function.

the etl.py command use the config file to decide with job submitter to use

Deploy a job using etl.py command: (examples/etl.py)

./etl.py -a deploy \
    -c <config-filename> \
    --build-dir <build-dir> \
    --deploy-dir <deploy-dir>

• -c <config-filename>: this option specify the config file to use for the deployment
• --build-dir <build-dir>: this option specify where to look for the build bits to deploy
• --deolpy-dir <deploy-dir>: this option specify what is the destination for the deployment

Run a job

./etl.py -a run \
    -c <config-filename> \
    --deploy-dir <deploy-dir> \
    --version <version> \
    --args <input-json-file>

• -c <config-filename>: this option specify the config file
• --build-dir <build-dir>: this option specify where to look for the build bits to run
• --version <version>: this option specify which version of the app to run
• --args <input-json-file>: optional parameter for input variable for the job. The <input-json-file> points to a json file, the value of the file will be passed to job's main function in input_args parameter. If this option is missing, the input_args will be set to {} when calling the main function of the job.
• It prints the return value of the main function of the job

Examples

• Build, deploy and run in local spark
• Build, deploy and run in spark cluster
• Build, deploy and run in OCI Dataflow