Integrated in Hadoop-trunk #611 (See http://hudson.zones.apache.org/hudson/job/Hadoop-trunk/611/)

I just committed this. Thanks, Sharad!

TestFileAppend2.testComplexAppend has failed on Hudson which is unrelated to this patch.

-1 overall. Here are the results of testing the latest attachment
http://issues.apache.org/jira/secure/attachment/12390255/3829_v4.patch
against trunk revision 696525.

+1 @author. The patch does not contain any @author tags.

+1 tests included. The patch appears to include 11 new or modified tests.

+1 javadoc. The javadoc tool did not generate any warning messages.

+1 javac. The applied patch does not increase the total number of javac compiler warnings.

+1 findbugs. The patch does not introduce any new Findbugs warnings.

-1 core tests. The patch failed core unit tests.

+1 contrib tests. The patch passed contrib unit tests.

Test results: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3293/testReport/
Findbugs warnings: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3293/artifact/trunk/build/test/findbugs/newPatchFindbugsWarnings.html
Checkstyle results: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3293/artifact/trunk/build/test/checkstyle-errors.html
Console output: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3293/console

This message is automatically generated.

ant test passed on my machine.

test-patch passed on my machine.

[exec] +1 overall.

[exec] +1 @author. The patch does not contain any @author tags.

[exec] +1 tests included. The patch appears to include 11 new or modified tests.

[exec] +1 javadoc. The javadoc tool did not generate any warning messages.

[exec] +1 javac. The applied patch does not increase the total number of javac compiler warnings.

[exec] +1 findbugs. The patch does not introduce any new Findbugs warnings.

Fixed an issue in ReduceTask#SkippingReduceValuesIterator.
Added more documentation to SkipBadRecords.
Made writing of skip records optional.
Incorporated Devaraj's offline comment to remove the SkipBadRecords.ENABLED flag as it is now redundant after the addition of MAPPER_MAX_SKIP_RECORDS/REDUCER_MAX_SKIP_GROUPS.

+1 overall. Here are the results of testing the latest attachment
http://issues.apache.org/jira/secure/attachment/12389656/3829_v3.patch
against trunk revision 692996.

+1 @author. The patch does not contain any @author tags.

+1 tests included. The patch appears to include 5 new or modified tests.

+1 javadoc. The javadoc tool did not generate any warning messages.

+1 javac. The applied patch does not increase the total number of javac compiler warnings.

+1 findbugs. The patch does not introduce any new Findbugs warnings.

+1 core tests. The patch passed core unit tests.

+1 contrib tests. The patch passed contrib unit tests.

Test results: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3205/testReport/
Findbugs warnings: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3205/artifact/trunk/build/test/findbugs/newPatchFindbugsWarnings.html
Checkstyle results: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3205/artifact/trunk/build/test/checkstyle-errors.html
Console output: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/3205/console

This message is automatically generated.

updated with the trunk.

fairly tested patch. few additional things:

• moved counters from Counters.java to SkipBadRecords.java as these are specific to skip feature.
• fixed hasNext in SortedRanges.SkipRangeIterator.
• skipped records are not written to HDFS, if it is a test attempt (test attempt is to figure out whether a range is good or bad. In this only records in test range are passed to the mapper/reducer, others are skipped).
• recordreader.next not called beyond the test range during test attempt.
• renamed failedRanges to skipRanges in Task.java
• added config params to hadoop-default

Attaching the working patch, while I continue to test.

This patch depends on the patch from HADOOP-153. Please apply 153_7.patch before applying this one.

The approach has been discussed earlier in HADOOP-153 as well. Here is the brief:

Defines user configurable MAPPER_MAX_SKIP_RECORDS/MAPPER_REDUCE_SKIP_RECORDS -> acceptable skipped records in the neighborhood of a bad record.
If skipped range is greater than this threshold, the task will try to narrow down the skipped range using a binary search kind of algorithm during task re-executions till this threshold is met or all task attempts are exhausted. The skipped range is divided into two halves and only one half get executed. Based on the subsequent failure, it figures out which half contains the bad record.