Document the "stream.non.zero.exit.status.is.failure" knob for streaming

• Key: HADOOP-3379
• Type: Improvement
• Status: Closed
• Resolution: Fixed
• Priority: Blocker
• Assignee: Amareshwari Sriramadasu
• Reporter: Arun C Murthy
• Component/s: documentation
• Affects Version/s: 0.17.0
• Fix Version/s: 0.18.0

Description

HADOOP-2057 added a useful feature: "stream.non.zero.exit.status.is.failure" to optionally treat non-zero exit code from streaming apps as fatal. We should document this on http://hadoop.apache.org/core/docs/current/streaming.html.

Comments

[Owen O'Malley]

I wonder if we should just make streaming always fail the task for non-zero exit codes.

[Amareshwari Sriramadasu]

I wonder if we should just make streaming always fail the task for non-zero exit codes.

+1

[Rick Cox]

It seems reasonable for the default to be true, as I think the discussion has shown users expect it.

Making it non-optional would remove the ability to run 'grep' and similar commands without a wrapper (assuming the pattern is not in every input split). Perhaps that's ok; we could just document a hint to wrap such commands in a shell script that always exits 0.

[Amareshwari Sriramadasu]

Here is patch doing the following:
1. Makes default value for stream.non.zero.exit.is.failure as true . Thus by default, streaming tasks exiting with non-zero exit status are considered to be failed tasks.
2. Adds the documentation to streaming.xml.

[Devaraj Das]

I think you should document the fact that people have to wrap the command in some cases as Rick pointed out.

[Amareshwari Sriramadasu]

Actually the patch uploaded has the configuration property stream.non.zero.exit.is.failure value as true.

If we want 'streaming always fail the task for non-zero exit codes', we can just remove the configuration property and document to wrap commands as Rick pointed. But doesn't this make applications incompatible from 0.17 to 0.18 ? Shouldnt we deprecate the configuration property for 0.18?
Thoughts?

I think making the configuration property stream.non.zero.exit.is.failure value as true makes sense.

[Hadoop QA]

-1 overall. Here are the results of testing the latest attachment
http://issues.apache.org/jira/secure/attachment/12383731/patch-3379.txt
against trunk revision 666056.

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

-1 tests included. The patch doesn't appear to include any new or modified tests.
Please justify why no tests are needed for this patch.

+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 release audit. The applied patch does not increase the total number of release audit 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/2630/testReport/
Findbugs warnings: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/2630/artifact/trunk/build/test/findbugs/newPatchFindbugsWarnings.html
Checkstyle results: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/2630/artifact/trunk/build/test/checkstyle-errors.html
Console output: http://hudson.zones.apache.org/hudson/job/Hadoop-Patch/2630/console

This message is automatically generated.

[Owen O'Malley]

+1

[Devaraj Das]

I just committed this. Thanks, Amareshwari!

[Hudson]

Integrated in Hadoop-trunk #520 (See http://hudson.zones.apache.org/hudson/job/Hadoop-trunk/520/)