[
https://issues.apache.org/jira/browse/HADOOP-9102?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13507466#comment-13507466
]
Bikas Saha commented on HADOOP-9102:
------------------------------------
Looks like I had written it this way to differentiate between task being
alive/dead vs an error in executing the command. This change means any error in
execution that results in a non-zero exist code would mean the task is assumed
to be dead. Not sure if thats correct but lets leave that for now since the
same thing could happen for Linux and the Java code is written with that
expectation too. This also makes the isAlive parsing redundant in Hadoop code.
+1
> winutils task isAlive does not return a non-zero exit code if the requested
> task is not alive
> ---------------------------------------------------------------------------------------------
>
> Key: HADOOP-9102
> URL: https://issues.apache.org/jira/browse/HADOOP-9102
> Project: Hadoop Common
> Issue Type: Bug
> Components: util
> Affects Versions: 1-win
> Reporter: Chris Nauroth
> Assignee: Chris Nauroth
> Attachments: HADOOP-9102-branch-1-win.1.patch
>
>
> Work on YARN-233 noted that winutils task isAlive does not return a non-zero
> exit code if the requested task is not alive. This is inconsistent with the
> equivalent check on Unix, kill -0, which uses a non-zero exit code to
> communicate status. By changing winutils task isAlive to return a non-zero
> exit code, we can make the error handling code on the Java side consistent,
> avoiding the need for additional if (WINDOWS) checks.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
