hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Owen O'Malley (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HADOOP-2007) Jobs use incorrect path to job.xml for different users
Date Mon, 08 Oct 2007 01:26:50 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-2007?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12533020
] 

Owen O'Malley commented on HADOOP-2007:
---------------------------------------

Please configure mapred.system.dir to be a constant and somewhere outside of /tmp. I personally
like to use something like /hadoop/system so that it is very clear that this is a framework
directory.

> Jobs use incorrect path to job.xml for different users
> ------------------------------------------------------
>
>                 Key: HADOOP-2007
>                 URL: https://issues.apache.org/jira/browse/HADOOP-2007
>             Project: Hadoop
>          Issue Type: Bug
>          Components: mapred
>    Affects Versions: 0.14.1
>         Environment: OS: Red Hat Enterprise Linux AS release 4 (Nahant Update 5) - AMD64
> JRE: java full version "1.6.0_02-b05"
>            Reporter: Ryan Lynch
>
> We run hadoop/hdfs under a generic username of 'hadoop'.  When submitting a job as user
'fred', the job.xml file gets created correctly as /tmp/*hadoop-fred*/mapred/system/<job_id>/job.xml,
but the JobInProgress constructor appears to use the incorrect path of /tmp/*hadoop-hadoop*/mapred/system/<job_id>/job.xml
for the copyToLocalFile operation.  This results in an exception because the job.xml file
does not exist under /tmp/hadoop-hadoop.
> I think the incorrect path gets created on line 133 in JobInProgress.java right before
the call to copyToLocalFile.  If I hardcode the correct path (just to test my theory) of "/tmp/hadoop-fred/mapred/system"
in place of the call to default_conf.getSystemDir(), the job kicks off and runs to completion
as the user 'fred' with hadoop/hdfs running as the user 'hadoop'.
> Here's a portion of the output when the job is submitted by user 'fred':
> [fred@hdm01 ~]$ hadoop --config /usr/local/hadoop/conf jar /usr/local/hadoop/current/hadoop-0.14.1-examples.jar
wordcount mytest/input/data output2
> 07/10/07 17:52:14 INFO mapred.FileInputFormat: Total input paths to process : 62
> org.apache.hadoop.ipc.RemoteException: java.io.IOException: /tmp/hadoop-hadoop/mapred/system/job_200710071751_0002/job.xml:
No such file or directory
>         at org.apache.hadoop.fs.FileUtil.copy(FileUtil.java:138)
>         at org.apache.hadoop.fs.FileSystem.copyToLocalFile(FileSystem.java:803)
>         at org.apache.hadoop.fs.FileSystem.copyToLocalFile(FileSystem.java:784)
>         at org.apache.hadoop.mapred.JobInProgress.<init>(JobInProgress.java:134)
>         at org.apache.hadoop.mapred.JobTracker.submitJob(JobTracker.java:1479)

-- 
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