[
https://issues.apache.org/jira/browse/HADOOP-3827?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12617002#action_12617002
]
Viraj Bhat commented on HADOOP-3827:
------------------------------------
Here are the output and error logs for the maps and reduces which can result
from this bug
--------------------------------------------------------------------------------------------------------------------------------------------------------------------
Logs of the output from killed map - "m_005937_0" with zero input and output
bytes to hdfs
--------------------------------------------------------------------------------------------------------------------------------------------------------------------
attempt_200807242354_0001_m_005937_0: No outputs to promote from
hdfs://ymachine.mydomain.com/myhome/dir/_temporary/_attempt_200807242354_0001_m_005937_0
2008-07-25 00:05:55,986 INFO org.apache.hadoop.mapred.TaskRunner: Task
'attempt_200807242354_0001_m_005937_0' done.
--------------------------------------------------------------------------------------------------------------------------------------------------------------------
Error on map-side
--------------------------------------------------------------------------------------------------------------------------------------------------------------------
Too many fetch-failures
Too many fetch-failures
Too many fetch-failures
--------------------------------------------------------------------------------------------------------------------------------------------------------------------
Logs of the output from killed reduce "attempt_200807242354_0001_r_000001_0 "
as a result of - map "m_005937_0" providing zero output bytes to the reducers
--------------------------------------------------------------------------------------------------------------------------------------------------------------------
2008-07-25 00:06:00,618 INFO org.apache.hadoop.mapred.ReduceTask: Shuffling 2
bytes (2 raw bytes) into RAM from attempt_200807242354_0001_m_005937_0
2008-07-25 00:06:00,618 INFO org.apache.hadoop.mapred.ReduceTask: Read 0 bytes
from map-output for attempt_200807242354_0001_m_005937_0
2008-07-25 00:06:00,618 WARN org.apache.hadoop.mapred.ReduceTask:
attempt_200807242354_0001_r_000001_0 copy failed:
attempt_200807242354_0001_m_005937_0 from mymachine.mydomain.com
2008-07-25 00:06:00,618 WARN org.apache.hadoop.mapred.ReduceTask:
java.io.IOException: Incomplete map output received for
attempt_200807242354_0001_m_005937_0 from
http://mymachine.mydomain.com:55279/mapOutput?job=job_200807242354_0001&map=attempt_200807242354_0001_m_005937_0&reduce=1
(0 instead of 2)
at
org.apache.hadoop.mapred.ReduceTask$ReduceCopier$MapOutputCopier.shuffleInMemory(ReduceTask.java:1248)
at
org.apache.hadoop.mapred.ReduceTask$ReduceCopier$MapOutputCopier.getMapOutput(ReduceTask.java:1093)
at
org.apache.hadoop.mapred.ReduceTask$ReduceCopier$MapOutputCopier.copyOutput(ReduceTask.java:983)
at
org.apache.hadoop.mapred.ReduceTask$ReduceCopier$MapOutputCopier.run(ReduceTask.java:932)
........
........
2008-07-25 00:06:37,696 INFO org.apache.hadoop.mapred.ReduceTask: Failed to
fetch map-output from attempt_200807242354_0001_m_005937_0 even after
MAX_FETCH_RETRIES_PER_MAP retries... reporting to the JobTracker
--------------------------------------------------------------------------------------------------------------------------------------------------------------------
> Jobs with empty map-outputs and intermediate compression fail
> -------------------------------------------------------------
>
> Key: HADOOP-3827
> URL: https://issues.apache.org/jira/browse/HADOOP-3827
> Project: Hadoop Core
> Issue Type: Bug
> Components: mapred
> Affects Versions: 0.18.0
> Reporter: Arun C Murthy
> Assignee: Arun C Murthy
> Priority: Blocker
> Fix For: 0.18.0
>
> Attachments: HADOOP-3827_0_20080724.patch
>
>
> The corner case where there are zero map-outputs doesn't pass the codec to
> the IFile.Writer leading to un-compressed data and subsequently failure on
> the reduce when it tries to decompress that data.
> The straight-forward fix is to pass the codec:
> {noformat}
> Writer<K, V> writer = new Writer<K, V>(job, finalOut,
> - keyClass, valClass, null);
> + keyClass, valClass, codec);
> {noformat}
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
