Submitting applications to a cluster

This recipe shows how to run an application on distributed clusters. An application is launched on a set of machines using an external service called a cluster manager. There is a wide variety of cluster managers such as Hadoop YARN, Apache Mesos, and Spark's own built-in standalone cluster manager. Spark provides a single tool for submitting jobs across all cluster managers, called spark-submit. Through various options, spark-submit can connect to different cluster managers and control how many resources your application gets.

Getting ready

To step through this recipe, you will need a running Spark cluster either in pseudo distributed mode or in one of the distributed modes, that is, standalone, YARN, or Mesos.

How to do it…

1. Let's create a word count application:

           package org.apache.spark.programs 
           object WordCount{ 
           def main(args:Array[String]) { 
           val conf = new SparkConf 
           conf.setAppName("WordCount") 
           val sc = new SparkContext(conf) 
           val input =     
           sc.parallelize(Array("this,is,a,ball","it,is,a,cat","john,is,
           in,town,hall")) 
           val words = input.flatMap{record => record.split(",")} 
           val wordPairs = words.map(word => (word,1)) 
           val wordCounts = wordPairs.reduceByKey{(a,b) => a+b} 
           val result = wordCounts.collect 
           println("Displaying the WordCounts:") 
           result.foreach(println) 
   

2. Submit the application to Spark's standalone cluster manager:

         spark-submit --class org.apache.spark.programs.WordCount --master 
         spark://master:7077 WordCount.jar 
   
   

3. Submit the application to YARN:

         spark-submit --class org.apache.spark.programs.WordCount --master 
         yarn WordCount.jar
   

4. Submit the application to Mesos:

         spark-submit --class org.apache.spark.programs.WordCount --master       
         mesos://mesos-master:5050 WordCount.jar
   

How it works…

When spark-submit is called with the --master flag as spark://master:7077 submits the application to Spark's standalone cluster. Invoking with the --master flag as yarn runs the application in the YARN cluster, whereas specifying the --master flag as mesos://mesos-master:5050 runs the application on Mesos cluster.

There's more…

Whenever spark-submit is invoked, it launches the driver program. This driver program contacts the cluster manager and requests resources to launch executors. Once the executors are launched by the cluster manager, the driver runs through the user application. It delegates the work to executors in the form of tasks. When the driver's main() method exits, it will terminate the executors and releases resources from the cluster manager. spark-submit provides various options as well to control specific details.

See also

For more information on submitting applications to a cluster and the various options provided by Spark-submit, please visit: http://spark.apache.org/docs/latest/submitting-applications.html. Also, for detailed information about the different cluster managers, please refer to the following:

Also, to learn in details about the different cluster managers, please refer:

• http://spark.apache.org/docs/latest/spark-standalone.html
• http://spark.apache.org/docs/latest/running-on-mesos.html
• http://spark.apache.org/docs/latest/running-on-yarn.html