Welcome to OGeek Q&A Community for programmer and developer-Open, Learning and Share
Welcome To Ask or Share your Answers For Others

Categories

0 votes
681 views
in Technique[技术] by (71.8m points)

scala - Spark runs out of memory when grouping by key

I am attempting to perform a simple transformation of common crawl data using Spark host on an EC2 using this guide, my code looks like this:

package ccminer

import org.apache.spark.SparkContext
import org.apache.spark.SparkContext._

object ccminer {
  val english = "english|en|eng"
  val spanish = "es|esp|spa|spanish|espanol"
  val turkish = "turkish|tr|tur|turc"
  val greek = "greek|el|ell"
  val italian = "italian|it|ita|italien"
  val all = (english :: spanish :: turkish :: greek :: italian :: Nil).mkString("|")

  def langIndep(s: String) = s.toLowerCase().replaceAll(all, "*")

  def main(args: Array[String]): Unit = {
    if (args.length != 3) {
      System.err.println("Bad command line")
      System.exit(-1)
    }

    val cluster = "spark://???"
    val sc = new SparkContext(cluster, "Common Crawl Miner",
      System.getenv("SPARK_HOME"), Seq("/root/spark/ccminer/target/scala-2.10/cc-miner_2.10-1.0.jar"))

    sc.sequenceFile[String, String](args(0)).map {
      case (k, v) => (langIndep(k), v)
    }
    .groupByKey(args(2).toInt)
    .filter {
      case (_, vs) => vs.size > 1
    }
    .saveAsTextFile(args(1))
  }
}

And I am running it with the command as follows:

sbt/sbt "run-main ccminer.ccminer s3n://aws-publicdatasets/common-crawl/parse-output/segment/1341690165636/textData-* s3n://parallelcorpus/out/ 2000"

But very quickly it fails with errors as follows

java.lang.OutOfMemoryError: Java heap space
at com.ning.compress.BufferRecycler.allocEncodingBuffer(BufferRecycler.java:59)
at com.ning.compress.lzf.ChunkEncoder.<init>(ChunkEncoder.java:93)
at com.ning.compress.lzf.impl.UnsafeChunkEncoder.<init>(UnsafeChunkEncoder.java:40)
at com.ning.compress.lzf.impl.UnsafeChunkEncoderLE.<init>(UnsafeChunkEncoderLE.java:13)
at com.ning.compress.lzf.impl.UnsafeChunkEncoders.createEncoder(UnsafeChunkEncoders.java:31)
at com.ning.compress.lzf.util.ChunkEncoderFactory.optimalInstance(ChunkEncoderFactory.java:44)
at com.ning.compress.lzf.LZFOutputStream.<init>(LZFOutputStream.java:61)
at org.apache.spark.io.LZFCompressionCodec.compressedOutputStream(CompressionCodec.scala:60)
at org.apache.spark.storage.BlockManager.wrapForCompression(BlockManager.scala:803)
at org.apache.spark.storage.BlockManager$$anonfun$5.apply(BlockManager.scala:471)
at org.apache.spark.storage.BlockManager$$anonfun$5.apply(BlockManager.scala:471)
at org.apache.spark.storage.DiskBlockObjectWriter.open(BlockObjectWriter.scala:117)
at org.apache.spark.storage.DiskBlockObjectWriter.write(BlockObjectWriter.scala:174)
at org.apache.spark.scheduler.ShuffleMapTask$$anonfun$runTask$1.apply(ShuffleMapTask.scala:164)
at org.apache.spark.scheduler.ShuffleMapTask$$anonfun$runTask$1.apply(ShuffleMapTask.scala:161)
at scala.collection.Iterator$class.foreach(Iterator.scala:727)
at scala.collection.AbstractIterator.foreach(Iterator.scala:1157)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:161)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:102)
at org.apache.spark.scheduler.Task.run(Task.scala:53)
at org.apache.spark.executor.Executor$TaskRunner$$anonfun$run$1.apply$mcV$sp(Executor.scala:213)
at org.apache.spark.deploy.SparkHadoopUtil.runAsUser(SparkHadoopUtil.scala:49)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:178)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
at java.lang.Thread.run(Thread.java:744)

So my basic question is, what is necessary to write a Spark task that can group by key with an almost unlimited amount of input without running out of memory?

See Question&Answers more detail:os

与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…
Welcome To Ask or Share your Answers For Others

1 Reply

0 votes
by (71.8m points)

The most common cause of java.lang.OutOfMemoryError exceptions in shuffle tasks (such as groupByKey, reduceByKey, etc.) is low level of parallelism.

You can increase default value by setting spark.default.parallelism property in configuration.


与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…
OGeek|极客中国-欢迎来到极客的世界,一个免费开放的程序员编程交流平台!开放,进步,分享!让技术改变生活,让极客改变未来! Welcome to OGeek Q&A Community for programmer and developer-Open, Learning and Share
Click Here to Ask a Question

...