Repo for Sahara Testing Frameworks
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
Vitaly Gridnev 881409ed7b add impl of SparkWordCount example 3 years ago
wordcountapp add impl of SparkWordCount example 3 years ago
NOTICE.txt Move default templates to framework repo 4 years ago
README.rst Move default templates to framework repo 4 years ago
sample_input.txt Move default templates to framework repo 4 years ago Move default templates to framework repo 4 years ago embedded jars: replace spark-example.jar 4 years ago
spark-wordcount.jar add impl of SparkWordCount example 3 years ago


Example Spark Job

This example contains the compiled classes for SparkPi extracted from the example jar distributed with Apache Spark version 1.3.1.

SparkPi example estimates Pi. It can take a single optional integer argument specifying the number of slices (tasks) to use.

Example spark-wordcount Job

spark-wordcount is a modified version of the WordCount example from Apache Spark. It can read input data from hdfs or swift container, then output the number of occurrences of each word to standard output or hdfs.

Launching wordcount job from Sahara UI

  1. Create a job binary that points to spark-wordcount.jar.
  2. Create a job template and set spark-wordcount.jar as the main binary of the job template.
  3. Create a Swift container with your input file. As example, you can upload sample_input.txt.
  4. Launch job:

    1. Put path to input file in args
    2. Put path to output file in args
    3. Fill the Main class input with the following class: sahara.edp.spark.SparkWordCount
    4. Put the following values in the job's configs: edp.spark.adapt_for_swift with value True, fs.swift.service.sahara.password with password for your username, and fs.swift.service.sahara.username with your username. These values are required for correct access to your input file, located in Swift.
    5. Execute the job. You will be able to view your output in hdfs.

Launching spark-kafka-example

  1. Create a cluster with Kafka Broker, ZooKeeper and Spark History Server. The Ambari plugin can be used for that purpose. Please, use your keypair during cluster creation to have the ability to ssh in instances with that processes. For simplicity, these services should located on same the node.
  2. Ssh to the node with the Kafka Broker service. Create a sample topic using the following command: path/ --create --zookeeper localhost:2181 \ --replication-factor 1 --partitions 1 --topic test-topic. Also execute path/ --broker-list \ localhost:6667 --topic test-topic and then put several messages in the topic. Please, note that you need to replace the values localhost and path with your own values.
  3. Download the Spark Streaming utils to the node with your Spark History Server from this URL: Now you are ready to launch your job from sahara UI.
  4. Create a job binary that points to Also you need to create a job that uses this job binary as a main binary.
  5. Execute the job with the following job configs: edp.spark.driver.classpath with a value that points to the utils downloaded during step 2. Also the job should be run with the following arguments: localhost:2181 as the first argument, test-topic as the second, and 30 as the third.
  6. Congratulations, your job was successfully launched!