Task not serializable exception while running apache spark job

Question

The following java program was written to experiment with apache spark.

The program tries to read a list of positive and negative words from a respective file, compare it to the master file and filter the results accordingly.

import java.io.Serializable;
import java.io.FileNotFoundException;
import java.io.File;
import java.util.*;
import java.util.Iterator;
import java.util.List;
import java.util.List;
import org.apache.spark.api.java.*;
import org.apache.spark.api.java.function.Function;

public class SimpleApp implements Serializable{
  public static void main(String[] args) {
    String logFile = "/tmp/master.txt"; // Should be some file on your system
    String positive = "/tmp/positive.txt"; // Should be some file on your system
    String negative = "/tmp/negative.txt"; // Should be some file on your system

    JavaSparkContext sc = new JavaSparkContext("local[4]", "Twitter Analyzer", "/home/welcome/Downloads/spark-1.1.0/", new String[]{"target/scala-2.10/Simple-assembly-0.1.0.jar"});

    JavaRDD<String> positiveComments = sc.textFile(logFile).cache();

    List<String> positiveList = GetSentiments(positive);
    List<String> negativeList= GetSentiments(negative);

    final Iterator<String> iterator = positiveList.iterator();
    int i = 0;
    while (iterator.hasNext())
    {
      JavaRDD<String> numAs = positiveComments.filter(new Function<String, Boolean>()
      {
        public Boolean call(String s)
        {
          return s.contains(iterator.next());
        }
      });

     numAs.saveAsTextFile("/tmp/output/"+ i);
     i++;
     }

  }

public static List<String> GetSentiments(String fileName) {
  List<String> input = new ArrayList<String>();
try
{
  Scanner sc = new Scanner(new File(fileName));

  while (sc.hasNextLine()) {
      input.add(sc.nextLine());
  }
}
catch (FileNotFoundException e){
    // do stuff here..
}
  return input;
}

}

The Following error is thrown while executing spark job,

 Exception in thread "main" org.apache.spark.SparkException: Task not serializable
    at org.apache.spark.util.ClosureCleaner$.ensureSerializable(ClosureCleaner.scala:166)
    at org.apache.spark.util.ClosureCleaner$.clean(ClosureCleaner.scala:158)
    at org.apache.spark.SparkContext.clean(SparkContext.scala:1242)
    at org.apache.spark.rdd.RDD.filter(RDD.scala:282)
    at org.apache.spark.api.java.JavaRDD.filter(JavaRDD.scala:78)
    at SimpleApp.main(SimpleApp.java:37)
    at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
    at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:57)
    at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
    at java.lang.reflect.Method.invoke(Method.java:606)
    at org.apache.spark.deploy.SparkSubmit$.launch(SparkSubmit.scala:328)
    at org.apache.spark.deploy.SparkSubmit$.main(SparkSubmit.scala:75)
    at org.apache.spark.deploy.SparkSubmit.main(SparkSubmit.scala)
Caused by: java.io.NotSerializableException: java.util.ArrayList$Itr
    at java.io.ObjectOutputStream.writeObject0(ObjectOutputStream.java:1183)
    at java.io.ObjectOutputStream.defaultWriteFields(ObjectOutputStream.java:1547)
    at java.io.ObjectOutputStream.writeSerialData(ObjectOutputStream.java:1508)
    at java.io.ObjectOutputStream.writeOrdinaryObject(ObjectOutputStream.java:1431)
    at java.io.ObjectOutputStream.writeObject0(ObjectOutputStream.java:1177)
    at java.io.ObjectOutputStream.defaultWriteFields(ObjectOutputStream.java:1547)
    at java.io.ObjectOutputStream.writeSerialData(ObjectOutputStream.java:1508)
    at java.io.ObjectOutputStream.writeOrdinaryObject(ObjectOutputStream.java:1431)
    at java.io.ObjectOutputStream.writeObject0(ObjectOutputStream.java:1177)
    at java.io.ObjectOutputStream.writeObject(ObjectOutputStream.java:347)
    at org.apache.spark.serializer.JavaSerializationStream.writeObject(JavaSerializer.scala:42)
    at org.apache.spark.serializer.JavaSerializerInstance.serialize(JavaSerializer.scala:73)
    at org.apache.spark.util.ClosureCleaner$.ensureSerializable(ClosureCleaner.scala:164)
    ... 12 more

Any pointers??

Answer 1

Some Java Facts

1. Any anonymous class defined inside a outer class has reference to the outer class.
2. If the anonymous class needs to be serialized it will compel you to make the outer class serialized.
3. Inside the lambda function if one uses a method of the enclosing class , the class needs to be serialized , if the lambda function is being serialized.

Some Facts about Spark.

1. On Same Executor multiple tasks can run at the same time in the same JVM as Tasks are spawned as threads in spark.
2. Any lambda, Anonymous Class used with the spark Transformation function (map, mapPartitions, keyBy , redudeByKey …) will be instantiated on driver, serialized and sent to the executor.
3. To serialize an object means to convert its state to a byte stream so that the byte stream can be reverted back into a copy of the object.
4. A Java object is serializable if its class or any of its super class implements either the java.io.Serializable interface or its subinterface, java.io.Externalizable, and all of its non-transient, non-static fields are serializable.

Rule of thumb to avoid serialization problem:

1. Avoid using anonymous class , instead use static classes as anonymous class will force you to have the outer class serialized.
2. Avoid using static variables as a work around for serialization issue , as Multiple Task can run inside the same JVM and the static instance might not be thread safe.
3. Use Transient variables to avoid serialization issue , you will have to initialize them inside the function call and not Constructor. As on driver the constructor will be called , on Executor it will de-serialize and for the object . only way to initialize is inside the function call .
4. Use Static class in place of anonymous class.
5. Religiously follow ” attaching implements Serializable ” only for the classes which only needs to be serialized
6. Inside a “lambda function” never refer to outclass method directly , as this will lead to serialization of outer class.
7. Make methods static if it needs to be used within Lambda function directly , else use Class::func() notion but not func() directly
8. Java Map<> doesn’t implement Serializable but HashMap does .
9. Be wise when deciding over using Braodcast vs Raw DataStructures. If you see a real benefit then only use Broadcast.

For a in depth understanding follow http://bytepadding.com/big-data/spark/understanding-spark-serialization/

Answer 2

When you create an anonymous class, the compiler does some stuff:

JavaRDD<String> numAs = positiveComments.filter(new Function<String, Boolean>()
      {
        public Boolean call(String s)
        {
          return s.contains(iterator.next());
        }
      });

It will be rewritten as:

JavaRDD<String> numAs = positiveComments.filter(new Function<String, Boolean>()
      {
        private Iterator<...> $iterator;
        public Boolean call(String s)
        {
          return s.contains($iterator.next());
        }
      });

This is why you can have a NotSerializableException because the Iterator is not serializable.

To avoid that, simply extract the result of next before:

String value = iterator.next();
JavaRDD<String> numAs = positiveComments.filter(new Function<String, Boolean>()
      {
        public Boolean call(String s)
        {
          return s.contains(value);
        }
      });