You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-dev@hadoop.apache.org by "Karl Anderson (JIRA)" <ji...@apache.org> on 2008/08/26 01:53:44 UTC
[jira] Created: (HADOOP-4022) mapper output truncated on
pseudo-distributed cluster
mapper output truncated on pseudo-distributed cluster
-----------------------------------------------------
Key: HADOOP-4022
URL: https://issues.apache.org/jira/browse/HADOOP-4022
Project: Hadoop Core
Issue Type: Bug
Affects Versions: 0.18.0, 0.17.0
Reporter: Karl Anderson
On a pseudo-distributed test run, I'm seeing truncated mapper output. I don't see this when running the same job on a real cluster managed with hadoop-ec2.
With a no-reducers streaming job, I'm getting several output files, each truncated to 325.24 KB, and several zero-length output files. With "cat" as my reducer, I get one output file, again truncated to 325.24 KB. This only happens in a pseudo-distributed Hadoop run - when I run the mapper on the command line (cat input_file | ./mapper.py | sort), I get the full output. Input splitting isn't a factor, my input file is small enough to fit in one input split (and truncated input would just break the mapper, not truncate its output).
My mapper is outputting the default key-value lines for streaming, with a tab separating the key and the value, and no newlines or tabs in the value.
Truncation happens in the middle of a line. My lines are very long, the lines themselves are over 325.24 KB. This isn't the most efficient use of Hadoop, I know, I'm just putting a job in my pipeline while I work on a better implementation.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Commented: (HADOOP-4022) mapper output truncated on
pseudo-distributed cluster
Posted by "Karl Anderson (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HADOOP-4022?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12625559#action_12625559 ]
Karl Anderson commented on HADOOP-4022:
---------------------------------------
This could be an architecture issue. The pseudo-distributed box is running MacOSX Leopard. The ec2 image is running a version of Fedora. However, these files are pretty small.
> mapper output truncated on pseudo-distributed cluster
> -----------------------------------------------------
>
> Key: HADOOP-4022
> URL: https://issues.apache.org/jira/browse/HADOOP-4022
> Project: Hadoop Core
> Issue Type: Bug
> Affects Versions: 0.17.0, 0.18.0
> Reporter: Karl Anderson
>
> On a pseudo-distributed test run, I'm seeing truncated mapper output. I don't see this when running the same job on a real cluster managed with hadoop-ec2.
> With a no-reducers streaming job, I'm getting several output files, each truncated to 325.24 KB, and several zero-length output files. With "cat" as my reducer, I get one output file, again truncated to 325.24 KB. This only happens in a pseudo-distributed Hadoop run - when I run the mapper on the command line (cat input_file | ./mapper.py | sort), I get the full output. Input splitting isn't a factor, my input file is small enough to fit in one input split (and truncated input would just break the mapper, not truncate its output).
> My mapper is outputting the default key-value lines for streaming, with a tab separating the key and the value, and no newlines or tabs in the value.
> Truncation happens in the middle of a line. My lines are very long, the lines themselves are over 325.24 KB. This isn't the most efficient use of Hadoop, I know, I'm just putting a job in my pipeline while I work on a better implementation.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.