Uploaded image for project: 'Spark'
  1. Spark
  2. SPARK-3958

Possible stream-corruption issues in TorrentBroadcast

    XMLWordPrintableJSON

Details

    • Bug
    • Status: Resolved
    • Blocker
    • Resolution: Fixed
    • 1.1.0, 1.2.0
    • None
    • Spark Core
    • None

    Description

      TorrentBroadcast deserialization sometimes fails with decompression errors, which are most likely caused by stream-corruption exceptions. For example, this can manifest itself as a Snappy PARSING_ERROR when deserializing a broadcasted task:

      14/10/14 17:20:55.016 DEBUG BlockManager: Getting local block broadcast_8
      14/10/14 17:20:55.016 DEBUG BlockManager: Block broadcast_8 not registered locally
      14/10/14 17:20:55.016 INFO TorrentBroadcast: Started reading broadcast variable 8
      14/10/14 17:20:55.017 INFO TorrentBroadcast: Reading broadcast variable 8 took 5.3433E-5 s
      14/10/14 17:20:55.017 ERROR Executor: Exception in task 2.0 in stage 8.0 (TID 18)
      java.io.IOException: PARSING_ERROR(2)
      	at org.xerial.snappy.SnappyNative.throw_error(SnappyNative.java:84)
      	at org.xerial.snappy.SnappyNative.uncompressedLength(Native Method)
      	at org.xerial.snappy.Snappy.uncompressedLength(Snappy.java:594)
      	at org.xerial.snappy.SnappyInputStream.readFully(SnappyInputStream.java:125)
      	at org.xerial.snappy.SnappyInputStream.readHeader(SnappyInputStream.java:88)
      	at org.xerial.snappy.SnappyInputStream.<init>(SnappyInputStream.java:58)
      	at org.apache.spark.io.SnappyCompressionCodec.compressedInputStream(CompressionCodec.scala:128)
      	at org.apache.spark.broadcast.TorrentBroadcast$.unBlockifyObject(TorrentBroadcast.scala:216)
      	at org.apache.spark.broadcast.TorrentBroadcast.readObject(TorrentBroadcast.scala:170)
      	at sun.reflect.GeneratedMethodAccessor92.invoke(Unknown Source)
      	at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
      	at java.lang.reflect.Method.invoke(Method.java:606)
      	at java.io.ObjectStreamClass.invokeReadObject(ObjectStreamClass.java:1017)
      	at java.io.ObjectInputStream.readSerialData(ObjectInputStream.java:1893)
      	at java.io.ObjectInputStream.readOrdinaryObject(ObjectInputStream.java:1798)
      	at java.io.ObjectInputStream.readObject0(ObjectInputStream.java:1350)
      	at java.io.ObjectInputStream.defaultReadFields(ObjectInputStream.java:1990)
      	at java.io.ObjectInputStream.readSerialData(ObjectInputStream.java:1915)
      	at java.io.ObjectInputStream.readOrdinaryObject(ObjectInputStream.java:1798)
      	at java.io.ObjectInputStream.readObject0(ObjectInputStream.java:1350)
      	at java.io.ObjectInputStream.readObject(ObjectInputStream.java:370)
      	at org.apache.spark.serializer.JavaDeserializationStream.readObject(JavaSerializer.scala:62)
      	at org.apache.spark.serializer.JavaSerializerInstance.deserialize(JavaSerializer.scala:87)
      	at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:164)
      	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:745)
      

      SPARK-3630 is an umbrella ticket for investigating all causes of these Kryo and Snappy deserialization errors. This ticket is for a more narrowly-focused exploration of the TorrentBroadcast version of these errors, since the similar errors that we've seen in sort-based shuffle seem to be explained by a different cause (see SPARK-3948).

      Attachments

        1. spark_ex.logs
          78 kB
          Vitaliy Migov

        Issue Links

          Activity

            People

              joshrosen Josh Rosen
              joshrosen Josh Rosen
              Votes:
              4 Vote for this issue
              Watchers:
              8 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved: