Why does Spark Streaming not read from Kafka topic?

Question

• Spark Streaming 1.6.0
• Apache Kafka 10.0.1

I use Spark Streaming to read from sample topic. The code runs with no errors or exceptions but I don't get any data on the console via print() method.

I checked to see if there are messages in the topic:

./bin/kafka-console-consumer.sh \
    --zookeeper ip-172-xx-xx-xxx:2181 \
    --topic sample \
    --from-beginning

And I am getting the messages:

message no. 1
message no. 2
message no. 3
message no. 4
message no. 5

The command to run the streaming job:

./bin/spark-submit \
    --conf "spark.executor.extraJavaOptions=-XX:+PrintGCDetails -XX:+PrintGCTimeStamps -XX:MaxDirectMemorySize=512m" \
    --jars /home/ubuntu/zifferlabs/target/ZifferLabs-1-jar-with-dependencies.jar \
    --class "com.zifferlabs.stream.SampleStream" \
    /home/ubuntu/zifferlabs/src/main/java/com/zifferlabs/stream/SampleStream.java

Here is the entire code:

import java.util.Arrays;
import java.util.HashMap;
import java.util.HashSet;
import java.util.Map;
import java.util.Set;

import org.apache.spark.SparkConf;
import org.apache.spark.api.java.function.Function;
import org.apache.spark.streaming.Duration;
import org.apache.spark.streaming.api.java.JavaDStream;
import org.apache.spark.streaming.api.java.JavaPairInputDStream;
import org.apache.spark.streaming.api.java.JavaStreamingContext;
import org.apache.spark.streaming.kafka.KafkaUtils;

import kafka.serializer.DefaultDecoder;
import kafka.serializer.StringDecoder;
import scala.Tuple2;

public class SampleStream {
  private static void processStream() {
    SparkConf conf = new SparkConf().setAppName("sampleStream")
            .setMaster("local[3]")
            .set("spark.serializer", "org.apache.spark.serializer.JavaSerializer")
            .set("spark.driver.memory", "2g").set("spark.streaming.blockInterval", "1000")
            .set("spark.driver.allowMultipleContexts", "true")
            .set("spark.scheduler.mode", "FAIR");

    JavaStreamingContext jsc = new JavaStreamingContext(conf, new Duration(Long.parseLong("2000")));

    String[] topics = "sample".split(",");
    Set<String> topicSet = new HashSet<String>(Arrays.asList(topics));
    Map<String, String> props = new HashMap<String, String>();
    props.put("metadata.broker.list", "ip-172-xx-xx-xxx:9092");
    props.put("kafka.consumer.id", "sample_con");
    props.put("group.id", "sample_group");
    props.put("zookeeper.connect", "ip-172-xx-xx-xxx:2181");
    props.put("zookeeper.connection.timeout.ms", "16000");

    JavaPairInputDStream<String, byte[]> kafkaStream =
      KafkaUtils.createDirectStream(jsc, String.class, byte[].class, StringDecoder.class,
                                    DefaultDecoder.class, props, topicSet);

    JavaDStream<String> data = kafkaStream.map(new Function<Tuple2<String,byte[]>, String>() {
      public String call(Tuple2<String, byte[]> arg0) throws Exception {
        System.out.println("$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$ value is: " + arg0._2().toString());
        return arg0._2().toString();
      }
    });

    data.print();

    System.out.println("Spark Streaming started....");
    jsc.checkpoint("/home/spark/sparkChkPoint");
    jsc.start();
    jsc.awaitTermination();
    System.out.println("Stopped Spark Streaming");
  }

  public static void main(String[] args) {
    processStream();
  }
}

Answer 1

I think you've got the code right, but the command line to execute it is incorrect.

You spark-submit the application as follows (formatting's mine + spark.executor.extraJavaOptions removed for simplicity):

./bin/spark-submit \
  --jars /home/ubuntu/zifferlabs/target/ZifferLabs-1-jar-with-dependencies.jar \
  --class "com.zifferlabs.stream.SampleStream" \
  /home/ubuntu/zifferlabs/src/main/java/com/zifferlabs/stream/SampleStream.java

I think it won't work since spark-submit submits your Java source code not the executable code.

Please spark-submit your application as follows:

./bin/spark-submit \
  --class "com.zifferlabs.stream.SampleStream" \
  /home/ubuntu/zifferlabs/target/ZifferLabs-1-jar-with-dependencies.jar

which is --class to define the "entry point" to your Spark application and the code with dependencies (as the only input argument for spark-submit).

Give it a shot and report back!