hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Srikanth Kakani (JIRA)" <j...@apache.org>
Subject [jira] Created: (HADOOP-1764) Inconsistancy between Mapper/Reducer book keeping
Date Wed, 22 Aug 2007 23:38:30 GMT
Inconsistancy between Mapper/Reducer book keeping
-------------------------------------------------

                 Key: HADOOP-1764
                 URL: https://issues.apache.org/jira/browse/HADOOP-1764
             Project: Hadoop
          Issue Type: Bug
          Components: mapred
         Environment: Related: HADOOP-1763 (Same environment)
Version: 0.15.0-dev, r565628
Compiled: Tue Aug 14 20:55:37 UTC 2007 by hadoopqa
1400 Nodes
            Reporter: Srikanth Kakani
            Priority: Blocker


Refer to HADOOP-1763

This occurs in that scenario once many job trackers are lost, reducers do not know where the
map outputs are present. They keep retrying the wrong node causing the reducers to run forever
without failures.

Relevant logs:
Reducer output:
2007-08-21 09:47:47,046 INFO org.apache.hadoop.mapred.ReduceTask: task_200708210155_0003_r_000006_2
Copying task_200708210155_0003_m_002598_0 output from node50
2007-08-21 09:47:53,643 WARN org.apache.hadoop.mapred.ReduceTask: task_200708210155_0003_r_000006_2
copy failed: task_200708210155_0003_m_002598_0 from node50
2007-08-21 09:47:53,643 WARN org.apache.hadoop.mapred.ReduceTask: java.io.FileNotFoundException:
http://wm511750.inktomisearch.com:50060/mapOutput?map=task_200708210155_0003_m_002598_0&reduce=6
	at sun.net.www.protocol.http.HttpURLConnection.getInputStream(HttpURLConnection.java:1243)
	at org.apache.hadoop.mapred.MapOutputLocation.getFile(MapOutputLocation.java:207)
	at org.apache.hadoop.mapred.ReduceTask$ReduceCopier$MapOutputCopier.copyOutput(ReduceTask.java:673)
	at org.apache.hadoop.mapred.ReduceTask$ReduceCopier$MapOutputCopier.run(ReduceTask.java:631)
2007-08-21 09:53:02,327 INFO org.apache.hadoop.mapred.ReduceTask: task_200708210155_0003_r_000006_2
Copying task_200708210155_0003_m_002598_0 output from node50
2007-08-21 09:53:02,333 WARN org.apache.hadoop.mapred.ReduceTask: task_200708210155_0003_r_000006_2
copy failed: task_200708210155_0003_m_002598_0 from node50
2007-08-21 09:53:02,333 WARN org.apache.hadoop.mapred.ReduceTask: java.io.FileNotFoundException:
http://node50:50060/mapOutput?map=task_200708210155_0003_m_002598_0&reduce=6
	at sun.net.www.protocol.http.HttpURLConnection.getInputStream(HttpURLConnection.java:1243)
	at org.apache.hadoop.mapred.MapOutputLocation.getFile(MapOutputLocation.java:207)
	at org.apache.hadoop.mapred.ReduceTask$ReduceCopier$MapOutputCopier.copyOutput(ReduceTask.java:673)
	at org.apache.hadoop.mapred.ReduceTask$ReduceCopier$MapOutputCopier.run(ReduceTask.java:631)
2007-08-21 09:57:33,899 INFO org.apache.hadoop.mapred.ReduceTask: task_200708210155_0003_r_000006_2
Copying task_200708210155_0003_m_002598_0 output from node50.inktomisearch.com.
2007-08-21 09:57:33,908 WARN org.apache.hadoop.mapred.ReduceTask: task_200708210155_0003_r_000006_2
copy failed: task_200708210155_0003_m_002598_0 from node50.inktomisearch.com
2007-08-21 09:57:33,908 WARN org.apache.hadoop.mapred.ReduceTask: java.io.FileNotFoundException:
http://node50:50060/mapOutput?map=task_200708210155_0003_m_002598_0&reduce=6
	at sun.net.www.protocol.http.HttpURLConnection.getInputStream(HttpURLConnection.java:1243)
	at org.apache.hadoop.mapred.MapOutputLocation.getFile(MapOutputLocation.java:207)
	at org.apache.hadoop.mapred.ReduceTask$ReduceCopier$MapOutputCopier.copyOutput(ReduceTask.java:673)
	at org.apache.hadoop.mapred.ReduceTask$ReduceCopier$MapOutputCopier.run(ReduceTask.java:631)
2007-08-21 10:00:56,337 INFO org.apache.hadoop.mapred.ReduceTask: task_200708210155_0003_r_000006_2
Copying task_200708210155_0003_m_002598_1 output from node75.inktomisearch.com.
2007-08-21 10:00:56,342 INFO org.apache.hadoop.mapred.ReduceTask: task_200708210155_0003_r_000006_2
done copying task_200708210155_0003_m_002598_1 output from node75
2007-08-21 10:02:17,486 INFO org.apache.hadoop.mapred.ReduceTask: task_200708210155_0003_r_000006_2
Ignoring obsolete copy result for Map Task: task_200708210155_0003_m_002598_0 from host: node50

Looking at TIP task_200708210155_0003_m_002598:

task_200708210155_0003_m_002598_0	node50	KILLED	0.00%		21-Aug-2007 09:38:49 	Lost task tracker
task_200708210155_0003_m_002598_1	node75	KILLED	0.00%		21-Aug-2007 11:22:42 	Lost task tracker
task_200708210155_0003_m_002598_2	node55	SUCCEEDED	100.00%	21-Aug-2007 11:22:46	21-Aug-2007
11:27:19 (4mins, 33sec) 	
task_200708210155_0003_m_002598_3	node49	KILLED	100.00%	21-Aug-2007 11:22:48	21-Aug-2007 11:27:48
(4mins, 59sec) 	Already completed TIP


Notes:
1. Even finally the reducer seems to fetch data from the incorrect TaskTracker, it is not
checking with the job tracker for the final/correct map output
2. It seems to retry more times and sleeps for longer time (looking at the interval of log
messages)
3. An obvious solution may be to go to the job tracker and directly get the correct map output
(I was able to get the correct map output from node55 using http, without any errors)

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message