[SPARK-10929] Tungsten fails to acquire memory writing to HDFS

• Type: Bug
• Status: Resolved
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: 1.5.0, 1.5.1
• Fix Version/s: 1.6.0
• Component/s: SQL
• Target Version/s: 1.6.0
• Assignee: Davies Liu
• Reporter: Franco
• Created: 05/Oct/15 20:01
• Updated: 03/Nov/15 15:34
• Resolved: 30/Oct/15 06:43

Description

We are executing 20 Spark SQL jobs in parallel using Spark Job Server and hitting the following issue pretty routinely.

40GB heap x 6 nodes. Have tried adjusting shuffle.memoryFraction from 0.2 -> 0.1 with no difference.

.16): org.apache.spark.SparkException: Task failed while writing rows.
        at org.apache.spark.sql.execution.datasources.DefaultWriterContainer.writeRows(WriterContainer.scala:250)
        at org.apache.spark.sql.execution.datasources.InsertIntoHadoopFsRelation$$anonfun$run$1$$anonfun$apply$mcV$sp$3.apply(InsertIntoHadoopFsRelation.scala:150)
        at org.apache.spark.sql.execution.datasources.InsertIntoHadoopFsRelation$$anonfun$run$1$$anonfun$apply$mcV$sp$3.apply(InsertIntoHadoopFsRelation.scala:150)
        at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:66)
        at org.apache.spark.scheduler.Task.run(Task.scala:88)
        at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:214)
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
        at java.lang.Thread.run(Thread.java:745)
Caused by: java.io.IOException: Unable to acquire 16777216 bytes of memory
        at org.apache.spark.util.collection.unsafe.sort.UnsafeExternalSorter.acquireNewPage(UnsafeExternalSorter.java:351)
        at org.apache.spark.util.collection.unsafe.sort.UnsafeExternalSorter.<init>(UnsafeExternalSorter.java:138)
        at org.apache.spark.util.collection.unsafe.sort.UnsafeExternalSorter.create(UnsafeExternalSorter.java:106)
        at org.apache.spark.sql.execution.UnsafeExternalRowSorter.<init>(UnsafeExternalRowSorter.java:68)
        at org.apache.spark.sql.execution.TungstenSort.org$apache$spark$sql$execution$TungstenSort$$preparePartition$1(sort.scala:146)
        at org.apache.spark.sql.execution.TungstenSort$$anonfun$doExecute$3.apply(sort.scala:169)
        at org.apache.spark.sql.execution.TungstenSort$$anonfun$doExecute$3.apply(sort.scala:169)
        at org.apache.spark.rdd.MapPartitionsWithPreparationRDD.prepare(MapPartitionsWithPreparationRDD.scala:50)
        at org.apache.spark.rdd.ZippedPartitionsBaseRDD$$anonfun$tryPrepareParents$1.applyOrElse(ZippedPartitionsRDD.scala:83)
        at org.apache.spark.rdd.ZippedPartitionsBaseRDD$$anonfun$tryPrepareParents$1.applyOrElse(ZippedPartitionsRDD.scala:82)
        at scala.runtime.AbstractPartialFunction.apply(AbstractPartialFunction.scala:33)
        at scala.collection.TraversableLike$$anonfun$collect$1.apply(TraversableLike.scala:278)
        at scala.collection.immutable.List.foreach(List.scala:318)
        at scala.collection.TraversableLike$class.collect(TraversableLike.scala:278)
        at scala.collection.AbstractTraversable.collect(Traversable.scala:105)
        at org.apache.spark.rdd.ZippedPartitionsBaseRDD.tryPrepareParents(ZippedPartitionsRDD.scala:82)
        at org.apache.spark.rdd.ZippedPartitionsRDD2.compute(ZippedPartitionsRDD.scala:97)
        at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:297)
        at org.apache.spark.rdd.RDD.iterator(RDD.scala:264)

I have tried setting spark.buffer.pageSize to both 1MB and 64MB (in spark-defaults.conf) and it makes no difference.

It also tries to acquire 33554432 bytes of memory in both cases.

Issue Links

• is related to: SPARK-10733 TungstenAggregation cannot acquire page after switching to sort-based
• is related to: SPARK-10783 Do track the pointer array in UnsafeInMemorySorter
• is related to: SPARK-10474 TungstenAggregation cannot acquire memory for pointer array after switching to sort-based
• links to: [Github] Pull Request #9241 (davies)