Hadoop Spill failure

Question

I'am currently working on a project using Hadoop 0.21.0, 985326 and a cluster of 6 worker nodes and a head node. Submitting a regular mapreduce job fails, but I have no idea why. Has anybody seen this exception before?

org.apache.hadoop.mapred.Child: Exception running child : java.io.IOException: Spill failed
    at org.apache.hadoop.mapred.MapTask$MapOutputBuffer.checkSpillException(MapTask.java:1379)
    at org.apache.hadoop.mapred.MapTask$MapOutputBuffer.access$200(MapTask.java:711)
    at org.apache.hadoop.mapred.MapTask$MapOutputBuffer$Buffer.write(MapTask.java:1193)
    at java.io.DataOutputStream.write(DataOutputStream.java:90)
    at org.apache.hadoop.io.Text.write(Text.java:290)
    at org.apache.hadoop.io.serializer.WritableSerialization$WritableSerializer.serialize(WritableSerialization.java:100)
    at org.apache.hadoop.io.serializer.WritableSerialization$WritableSerializer.serialize(WritableSerialization.java:84)
    at org.apache.hadoop.mapred.MapTask$MapOutputBuffer.collect(MapTask.java:967)
    at org.apache.hadoop.mapred.MapTask$NewOutputCollector.write(MapTask.java:583)
    at org.apache.hadoop.mapreduce.task.TaskInputOutputContextImpl.write(TaskInputOutputContextImpl.java:92)
    at org.apache.hadoop.mapreduce.lib.map.WrappedMapper$Context.write(WrappedMapper.java:111)
    at be.ac.ua.comp.ronny.riki.invertedindex.FilteredInvertedIndexBuilder$Map.map(FilteredInvertedIndexBuilder.java:113)
    at be.ac.ua.comp.ronny.riki.invertedindex.FilteredInvertedIndexBuilder$Map.map(FilteredInvertedIndexBuilder.java:1)
    at org.apache.hadoop.mapreduce.Mapper.run(Mapper.java:144)
    at org.apache.hadoop.mapred.MapTask.runNewMapper(MapTask.java:652)
    at org.apache.hadoop.mapred.MapTask.run(MapTask.java:328)
    at org.apache.hadoop.mapred.Child$4.run(Child.java:217)
    at java.security.AccessController.doPrivileged(Native Method)
    at javax.security.auth.Subject.doAs(Subject.java:396)
    at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:742)
    at org.apache.hadoop.mapred.Child.main(Child.java:211)
Caused by: java.lang.RuntimeException: java.lang.NoSuchMethodException: org.apache.hadoop.io.ArrayWritable.<init>()
    at org.apache.hadoop.util.ReflectionUtils.newInstance(ReflectionUtils.java:123)
    at org.apache.hadoop.io.serializer.WritableSerialization$WritableDeserializer.deserialize(WritableSerialization.java:68)
    at org.apache.hadoop.io.serializer.WritableSerialization$WritableDeserializer.deserialize(WritableSerialization.java:44)
    at org.apache.hadoop.mapreduce.task.ReduceContextImpl.nextKeyValue(ReduceContextImpl.java:145)
    at org.apache.hadoop.mapreduce.task.ReduceContextImpl.nextKey(ReduceContextImpl.java:121)
    at org.apache.hadoop.mapreduce.lib.reduce.WrappedReducer$Context.nextKey(WrappedReducer.java:291)
    at org.apache.hadoop.mapreduce.Reducer.run(Reducer.java:168)
    at org.apache.hadoop.mapred.Task$NewCombinerRunner.combine(Task.java:1432)
    at org.apache.hadoop.mapred.MapTask$MapOutputBuffer.sortAndSpill(MapTask.java:1457)
    at org.apache.hadoop.mapred.MapTask$MapOutputBuffer.access$600(MapTask.java:711)
    at org.apache.hadoop.mapred.MapTask$MapOutputBuffer$SpillThread.run(MapTask.java:1349)
Caused by: java.lang.NoSuchMethodException: org.apache.hadoop.io.ArrayWritable.<init>()
    at java.lang.Class.getConstructor0(Class.java:2706)
    at java.lang.Class.getDeclaredConstructor(Class.java:1985)
    at org.apache.hadoop.util.ReflectionUtils.newInstance(ReflectionUtils.java:117)
    ... 10 more

Currently, I'm experimenting with some configuration parameters hoping that this error disappears, but until now this was unsuccessful. The configuration parameters I'm tweaking are:

• mapred.map.tasks = 60
• mapred.reduce.tasks = 12
• Job.MAP_OUTPUT_COMPRESS (or mapreduce.map.output.compress) = true
• Job.IO_SORT_FACTOR (or mapreduce.task.io.sort.factor) = 10
• Job.IO_SORT_MB (or mapreduce.task.io.sort.mb) = 256
• Job.MAP_JAVA_OPTS (or mapreduce.map.java.opts) = "-Xmx256" or "-Xmx512"
• Job.REDUCE_JAVA_OPTS (or mapreduce.reduce.java.opts) = "-Xmx256" or "-Xmx512"

Can anybody explain why the exception above occurs? And how to avoid it? Or just a short explanation what the hadoop spill operation implies?

Solution 1

Ok, all problems are solved.

The Map-Reduce serialization operation needs intern a default constructor for org.apache.hadoop.io.ArrayWritable.
Hadoops implementation didn't provide a default constructor for ArrayWritable.
That's why the java.lang.NoSuchMethodException: org.apache.hadoop.io.ArrayWritable.() was thrown and caused the weird spill exception.

A simple wrapper made ArrayWritable really writable and fixed it! Strange that Hadoop did not provide this.

Solution 2

This problem came up for me when the output of one of my map jobs produced a tab character ("\t") or newline character ("\r" or "\n") - Hadoop doesn't handle this well and fails. I was able to solve this using this piece of Python code:

if "\t" in output:
  output = output.replace("\t", "")
if "\r" in output:
  output = output.replace("\r", "")
if "\n" in output:
  output = output.replace("\n", "")

You may have to do something else for your app.

Comments

[MrGomez]

See my answer here on why this is the case: stackoverflow.com/questions/4386781/…