'Connection Refused' error while running Spark Streaming on local machine

Question

I know there are many threads already on 'spark streaming connection refused' issues. But most of these are in Linux or at least pointing to HDFS. I am running this on my local laptop with Windows.

I am running a very simple basic Spark streaming standalone application, just to see how the streaming works. Not doing anything complex here:-

import org.apache.spark.streaming.Seconds
import org.apache.spark.streaming.StreamingContext
import org.apache.spark.SparkConf

object MyStream 
{
    def main(args:Array[String]) 
    {
        val sc = new StreamingContext(new SparkConf(),Seconds(10))
        val mystreamRDD = sc.socketTextStream("localhost",7777)
        mystreamRDD.print()
        sc.start()
        sc.awaitTermination()
    }
}

I am getting the following error:-

2015-07-25 18:13:07 INFO  ReceiverSupervisorImpl:59 - Starting receiver
2015-07-25 18:13:07 INFO  ReceiverSupervisorImpl:59 - Called receiver onStart
2015-07-25 18:13:07 INFO  SocketReceiver:59 - Connecting to localhost:7777
2015-07-25 18:13:07 INFO  ReceiverTracker:59 - Registered receiver for      stream 0 from 192.168.19.1:11300
2015-07-25 18:13:08 WARN  ReceiverSupervisorImpl:92 - Restarting receiver     with delay 2000 ms: Error connecting to localhost:7777
java.net.ConnectException: Connection refused

I have tried using different port numbers, but it doesn't help. So it keeps retrying in loop and keeps on getting same error. Does anyone have an idea?

Answer 1

Make sure to start the netcat or the port connection before you run the program. nc -lk 8080

Answer 2

Within the code for socketTextStream, Spark creates an instance of SocketInputDStream which uses java.net.Socket https://github.com/apache/spark/blob/master/streaming/src/main/scala/org/apache/spark/streaming/dstream/SocketInputDStream.scala#L73

java.net.Socket is a client socket, which means it is expecting there to be a server already running at the address and port you specify. Unless you have some service running a server on port 7777 of your local machine, the error you are seeing is as expected.

To see what I mean, try the following (you may not need to set master or appName in your environment).

import org.apache.spark.streaming.Seconds
import org.apache.spark.streaming.StreamingContext
import org.apache.spark.SparkConf

object MyStream
{
  def main(args:Array[String])
  {
    val sc = new StreamingContext(new SparkConf().setMaster("local").setAppName("socketstream"),Seconds(10))
    val mystreamRDD = sc.socketTextStream("bbc.co.uk",80)
    mystreamRDD.print()
    sc.start()
    sc.awaitTermination()
  }
}

This doesn't return any content because the app doesn't speak HTTP to the bbc website but it does not get a connection refused exception.

To run a local server when on linux, I would use netcat with a simple command such as

cat data.txt | ncat -l -p 7777

I'm not sure what your best approach is in Windows. You could write another application which listens as a server on that port and sends some data.