[
https://issues.apache.org/jira/browse/CRUNCH-360?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13917600#comment-13917600
]
Gabriel Reid commented on CRUNCH-360:
-------------------------------------
Thanks for trying that out, and sorry it didn't work out. I put some more time
into debugging this issue this evening.
About the HADOOP_USER_CLASSPATH_FIRST and mapreduce.job.user.classpath.first
config properties: I totally overlooked the fact that those won't work in the
case of running with the local job tracker (which is what you're running), as
everything runs in a single JVM with the initial classpath when the tool is
started.
I tried running your example with a clean download of hadoop 2.2.0 with
avro-1.7.4.jar under share/hadoop/common/lib replaced with avro-1.7.6.jar, and
I no longer got the exact issue that you were running into. However, I got
another issue, with appears to be another bug with Avro.
I've logged the Avro bug and filed a patch under AVRO-1473. I also tried your
test code again with the patched Avro snapshot placed under
share/hadoop/common/lib, and both the namespaced and non-namespaced versions
work.
To get around this issue without using namespaces in Avro means that you'll
need to use a patched/newer version of Avro for now. The
HADOOP_USER_CLASSPATH_FIRST env variable and mapreduce.job.user.classpath.first
config property should help you out if you're running on an actual cluster, but
running in the local job tracker will require replacing the Avro jar manually
in your hadoop installation.
> GenericData.Record avro records without schema namespace gets implicit
> namespace"crunch"
> ----------------------------------------------------------------------------------------
>
> Key: CRUNCH-360
> URL: https://issues.apache.org/jira/browse/CRUNCH-360
> Project: Crunch
> Issue Type: Bug
> Reporter: Magnus Runesson
> Attachments: ImplicitNamespaceSchemaIT.java, crunch-example.tar.gz
>
>
> When having avroschema without namespace crunch implicit adds the namespace
> "crunch" when working with the records. Unfortunately this is not happening
> to the schema when reading an avrofile with At.avroFile(Path path,
> Configuration conf). The schema still has no namespace.
> In my job it ends up that my job fails looking up in the schema with the
> following error:
> The job uses Avro GenericData.Record and gets the schema from the avro file
> that is read.
> 2014-02-27 09:58:14,236 WARN [main] org.apache.hadoop.mapred.YarnChild:
> Exception running child : org.apache.avro.UnresolvedUnionException: Not in
> union
> [{"type":"record","name":"UsernameToUserId","namespace":"crunch","fields":[{"name":"username","type":["string","null"]},{"name":"user_id","type":["string","null"]}]},"null"]:
> {"username": "XXXXXXXXXX", "user_id": "XXXXXXXXXXXXXXXXXX"}
> at
> org.apache.avro.generic.GenericData.resolveUnion(GenericData.java:561)
> at
> org.apache.avro.generic.GenericDatumWriter.resolveUnion(GenericDatumWriter.java:144)
> at
> org.apache.avro.generic.GenericDatumWriter.write(GenericDatumWriter.java:71)
> at
> org.apache.avro.generic.GenericDatumWriter.writeRecord(GenericDatumWriter.java:106)
> at
> org.apache.avro.generic.GenericDatumWriter.write(GenericDatumWriter.java:66)
> at
> org.apache.avro.generic.GenericDatumWriter.write(GenericDatumWriter.java:58)
> at
> org.apache.crunch.types.avro.SafeAvroSerialization$AvroWrapperSerializer.serialize(SafeAvroSerialization.java:128)
> at
> org.apache.crunch.types.avro.SafeAvroSerialization$AvroWrapperSerializer.serialize(SafeAvroSerialization.java:113)
> at
> org.apache.hadoop.mapred.MapTask$MapOutputBuffer.collect(MapTask.java:1135)
> at
> org.apache.hadoop.mapred.MapTask$NewOutputCollector.write(MapTask.java:691)
> at
> org.apache.hadoop.mapreduce.task.TaskInputOutputContextImpl.write(TaskInputOutputContextImpl.java:89)
> at
> org.apache.hadoop.mapreduce.lib.map.WrappedMapper$Context.write(WrappedMapper.java:112)
> at
> org.apache.crunch.impl.mr.emit.OutputEmitter.emit(OutputEmitter.java:41)
> at org.apache.crunch.MapFn.process(MapFn.java:34)
> at org.apache.crunch.impl.mr.run.RTNode.process(RTNode.java:98)
> at
> org.apache.crunch.impl.mr.emit.IntermediateEmitter.emit(IntermediateEmitter.java:56)
> at org.apache.crunch.MapFn.process(MapFn.java:34)
> at org.apache.crunch.impl.mr.run.RTNode.process(RTNode.java:98)
> at
> org.apache.crunch.impl.mr.emit.IntermediateEmitter.emit(IntermediateEmitter.java:56)
> at org.apache.crunch.MapFn.process(MapFn.java:34)
> at org.apache.crunch.impl.mr.run.RTNode.process(RTNode.java:98)
> at
> org.apache.crunch.impl.mr.emit.IntermediateEmitter.emit(IntermediateEmitter.java:56)
> at org.apache.crunch.MapFn.process(MapFn.java:34)
> at org.apache.crunch.impl.mr.run.RTNode.process(RTNode.java:98)
> at org.apache.crunch.impl.mr.run.RTNode.process(RTNode.java:109)
> at org.apache.crunch.impl.mr.run.CrunchMapper.map(CrunchMapper.java:60)
> at org.apache.hadoop.mapreduce.Mapper.run(Mapper.java:145)
> at org.apache.hadoop.mapred.MapTask.runNewMapper(MapTask.java:763)
> at org.apache.hadoop.mapred.MapTask.run(MapTask.java:339)
> at org.apache.hadoop.mapred.YarnChild$2.run(YarnChild.java:162)
> 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:1491)
> at org.apache.hadoop.mapred.YarnChild.main(YarnChild.java:157)
> Either implicit namespace "crunch" should not be added anywhere. Or it must
> be added, if no namespace provided, when reading schema from the avro file i
> At.java:
> public static SourceTarget<GenericData.Record> avroFile(Path path,
> Configuration conf) {
> return avroFile(path, Avros.generics(From.getSchemaFromPath(path, conf)));
> }
> I use on HEAD in master.
>
--
This message was sent by Atlassian JIRA
(v6.2#6252)
