BroadCast Variable publish in Spark Program

Question

In the spark - java program I need to read a config file and populate a HashMap , which I need to publish as broadcast variable so that it will be available across all the datanodes .

I need to get the value of this broadcast variable in the CustomInputFormat class which is going to run in the datanodes . How can i specify in my CustomInputFormat class to get value from the specific broadcast variable since the broadcast variable is declared in my driver program ?

I am adding some code to explain it in more :

In this scenario1 I am using it in Driver Program itself ie the variable is used in the same class : Here I can use Broadcat.value() method

> final Broadcast<String[]> signPrefixes =
> sc.broadcast(loadCallSignTable());
>     JavaPairRDD<String, Integer> countryContactCounts = contactCounts.mapToPair(
>       new PairFunction<Tuple2<String, Integer>, String, Integer> (){
>         public Tuple2<String, Integer> call(Tuple2<String, Integer> callSignCount) {
>           String sign = callSignCount._1();
>           String country = lookupCountry(sign, signPrefixes.value());
>           return new Tuple2(country, callSignCount._2());
>         }}).reduceByKey(new SumInts());

In the scenario 2 I am going to use the Broadcast Variable inside my Custom Input Format class :

Driver Program :

> final JavaSparkContext sc=    new
> JavaSparkContext(sConf.setAppName("ParserSpark").setMaster("yarn-cluster"));
> Broadcast<int[]> broadcastVar = sc.broadcast(new int[] {1, 2, 3});
> 
> JavaPairRDD<NullWritable, ArrayList<Record>> baseRDD =
> sc.newAPIHadoopFile(args[2], InputFormat.class, NullWritable.class,
> ArrayList.class, conf);

InputFormat.class

> public class InputFormat extends  FileInputFormat {
> 
>   @Override   public RecordReader<NullWritable, ArrayList<Record>> 
>   createRecordReader(InputSplit split,            TaskAttemptContext context)
> throws IOException,           InterruptedException{
>       //I want to get the Broadcast Variable Here -- How will I do it 
>       
>         RecordReader reader = new RecordReader();         reader.initialize(split, context);      return reader;  }   @Override
>   protected boolean isSplitable(JobContext context, Path file) {
>       return false;    } }

Answer 1

I ran into this myself recently. Ended being rather simple actually (after a few hours and then a... a Ha!)

Create a new Configuration, set your vars, and pass it to a slightly different implementation of the newAPIHadoopFile function.

From the driver program (using Scala here):

val myConf = new Configuration();
    myConf.set("var1", v1)
    myConf.set("var2", v2)
    myConf.set("var3", v3)

val yourFile = sc.newAPIHadoopFile("yourFilePath", classOf[MyFileInputFormat],classOf[org.apache.hadoop.io.Text], classOf[org.apache.hadoop.io.DoubleWritable],myConf)

From your InputFormat or InputReader..or wherever you have a context (Java this time)

context.getConfiguration().get("var1");

or maybe

job.getConfiguration().get("var2");

Answer 2

You would create the broadcast var on the driver w/ val bcVariable = sc.broadcast(myVariableToBroadcast) and access it later w/ bcVariable.value