[Hadoop-common-user] ClassCastException with LineRecordReader (hadoop release version 0.21.0)

Hmm.. This has been fixed in MAPREDUCE-1905, in 0.21.1

Thanks
Amareshwari
On 4/21/11 7:27 AM, "Claus Stadler" wrote:

Hi,

I guess I am not the first one to see the following exception when
trying to initialize a LineRecordReader. However, so far I could't
figure out a workaround for this problem.

I saw that this problem was fixed in the svn, but when I checked out one
of the 0.23.0 versions (I can't remember which) I couldn't get the
servers started. (Not sure if that was due to the revision, or me
messing up the setup).

So maybe someone can point me to a known workaround for this problem or
at least hinting a revision that other people got to work?

Kind regards,
Claus


ps: Essentially I am just doing this:

public static class MyInputFormat extends
FileInputFormat<Text, ByteWritable>
{
@Override
public RecordReader<Text, ByteWritable>
createRecordReader(InputSplit inputSplit, TaskAttemptContext
taskAttemptContext)
throws IOException, InterruptedException
{
MyRecordReader result = new MyRecordReader();
result.initialize(inputSplit, taskAttemptContext);
return result;
}
}

public static class MyRecordReader extends
RecordReader<Text, ByteWritable>
{
LineRecordReader myReader = new LineRecordReader();
...
@Override
public void initialize(InputSplit inputSplit,
TaskAttemptContext taskAttemptContext)
throws IOException, InterruptedException
{
myReader.initialize(inputSplit, taskAttemptContext); //
EXCEPTION THROWN HERE
}
...
}

job.setInputFormatClass(MyInputFormat.class);


The exception is:

java.lang.Exception: java.lang.ClassCastException:
org.apache.hadoop.mapreduce.task.TaskAttemptContextImpl cannot be cast
to org.apache.hadoop.mapreduce.MapContext
at
org.apache.hadoop.mapred.LocalJobRunner$Job.run(LocalJobRunner.java:371)
~[hadoop-mapred-0.21.0.jar:na]
java.lang.ClassCastException:
org.apache.hadoop.mapreduce.task.TaskAttemptContextImpl cannot be cast
to org.apache.hadoop.mapreduce.MapContext
at
org.apache.hadoop.mapreduce.lib.input.LineRecordReader.initialize(LineRecordReader.java:75)
~[hadoop-mapred-0.21.0.jar:na]
(rest of the output is within my code)

Hi,


Thank you for the reply, however now I have the question on what the
recommended way is to get hadoop working with this fix?
Is there a documentation for this?

So some of my questions right now are:
. Does the svn head revision usually work?
. If not, is there a specific revision that is known to work (at least
in regard to this bug)?
. Or do I need to apply the patch?
. Do I need to do some special post-configuration after checking the
three sub-projects (common, mapred, hdfs) out from svn?
. Is there an estimate on the next release?

Kind regards,
Claus

Hi,

Sorry to bother you again, but could someone please give me advice on
how to set up hadoop with a fixed LineRecordReader?
(If I need to checkout from SVN, how do I get those three subprojects
combined und the bash scripts working? I couldn't find a doc for this yet.)

Kind regards,
Claus