qcri/Arabesque

Encounter kyro serialization error when calculating maximum cliques

Opened this issue · 2 comments

When I launch maximum clique computation on spark, I encounter the following errors(I bump spark version from 2.0.0 to 2.1.0):

Exception,java.lang.NullPointerException
Serialization trace:
vertexPositions (io.arabesque.pattern.JBlissPattern),[Ljava.lang.StackTraceElement;@106a45fa,com.esotericsoftware.kryo.KryoException: java.lang.NullPointerException
Serialization trace:
vertexPositions (io.arabesque.pattern.JBlissPattern)
at com.esotericsoftware.kryo.serializers.ObjectField.read(ObjectField.java:144)
at com.esotericsoftware.kryo.serializers.FieldSerializer.read(FieldSerializer.java:551)
at com.esotericsoftware.kryo.Kryo.readClassAndObject(Kryo.java:790)
at com.twitter.chill.Tuple2Serializer.read(TupleSerializers.scala:41)
at com.twitter.chill.Tuple2Serializer.read(TupleSerializers.scala:33)
at com.esotericsoftware.kryo.Kryo.readClassAndObject(Kryo.java:790)
at org.apache.spark.serializer.KryoDeserializationStream.readObject(KryoSerializer.scala:244)
at org.apache.spark.serializer.DeserializationStream.readKey(Serializer.scala:157)
at org.apache.spark.serializer.DeserializationStream$$anon$2.getNext(Serializer.scala:189)
at org.apache.spark.serializer.DeserializationStream$$anon$2.getNext(Serializer.scala:186)
at org.apache.spark.util.NextIterator.hasNext(NextIterator.scala:73)
at scala.collection.Iterator$$anon$12.hasNext(Iterator.scala:438)
at scala.collection.Iterator$$anon$11.hasNext(Iterator.scala:408)
at org.apache.spark.util.CompletionIterator.hasNext(CompletionIterator.scala:32)
at org.apache.spark.InterruptibleIterator.hasNext(InterruptibleIterator.scala:39)
at org.apache.spark.util.collection.ExternalAppendOnlyMap.insertAll(ExternalAppendOnlyMap.scala:154)
at org.apache.spark.Aggregator.combineCombinersByKey(Aggregator.scala:50)
at org.apache.spark.shuffle.BlockStoreShuffleReader.read(BlockStoreShuffleReader.scala:85)
at org.apache.spark.rdd.ShuffledRDD.compute(ShuffledRDD.scala:109)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:323)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:287)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:38)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:323)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:287)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:96)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:53)
at org.apache.spark.scheduler.Task.run(Task.scala:99)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:282)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at java.lang.Thread.run(Thread.java:748)
Caused by: java.lang.NullPointerException
at com.koloboke.collect.impl.hash.MutableLHashParallelKVIntIntMapGO.put(MutableLHashParallelKVIntIntMapGO.java:474)
at com.koloboke.collect.impl.hash.MutableLHashParallelKVIntIntMapGO.put(MutableLHashParallelKVIntIntMapGO.java:45)
at com.esotericsoftware.kryo.serializers.MapSerializer.read(MapSerializer.java:162)
at com.esotericsoftware.kryo.serializers.MapSerializer.read(MapSerializer.java:39)
at com.esotericsoftware.kryo.Kryo.readObject(Kryo.java:708)
at com.esotericsoftware.kryo.serializers.ObjectField.read(ObjectField.java:125)
... 30 more
,Some(org.apache.spark.ThrowableSerializationWrapper@5d417398),Vector(AccumulableInfo(1974,Some(internal.metrics.executorRunTime),Some(354),None,true,true,None), AccumulableInfo(1976,Some(internal.metrics.resultSize),Some(0),None,true,true,None), AccumulableInfo(1977,Some(internal.metrics.jvmGCTime),Some(27),None,true,true,None)),Vector(LongAccumulator(id: 1974, name: Some(internal.metrics.executorRunTime), value: 354), LongAccumulator(id: 1976, name: Some(internal.metrics.resultSize), value: 0), LongAccumulator(id: 1977, name: Some(internal.metrics.jvmGCTime), value: 27))),org.apache.spark.scheduler.TaskInfo@ab7ac0e,org.apache.spark.executor.TaskMetrics@2c4b63fc)

Does the spark version of Arabesque work with spark 2.1.0?

Arabesque is tested only with Spark 2.0. We faced several stability and memory issues when we tried upgrading to Spark 2.1+ (this NPE is one of them). We strongly recommend sticking to Spark 2.0 in order to get the expected outcomes.

I fixed the issue by forcing Spark to use Java serializer. In Spark 2.1+ kyro serializer is used by default in some of the shuffling stages, which destroys the serialization process. In the implementation most classes inherits Java serializer.