[ 
https://issues.apache.org/jira/browse/HADOOP-2092?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12557354#action_12557354
 ] 

Christian Kunz commented on HADOOP-2092:
----------------------------------------

The patch did not work for me. I tested the patch with a pipes application not 
emitting anything, just reading from DFS using libhdfs, and the application 
continued long after killing the task till it tried to shutdown the socket.

Another major headache of the current implementation is that the 'PING' enum 
value replaced the enum value for 'DONE' which increased by 1. This way the 
release process of applications needs to be sync'd up with Hadoop releases. It 
proved so inconvenient that I switched the two enums around in my patch.

On Unix checking for (getppid() == 1) works nicely. If there is need for pipes 
on Windows (is there?), there  might be no non-ugly way to retrieve the parent 
process, so I would suggest to use a ping from the parent to the child, and the 
child exits when there is no ping for a certain period.

> Pipes C++ task does not die even if the Java tasks die
> ------------------------------------------------------
>
>                 Key: HADOOP-2092
>                 URL: https://issues.apache.org/jira/browse/HADOOP-2092
>             Project: Hadoop
>          Issue Type: Bug
>          Components: pipes
>            Reporter: Amareshwari Sri Ramadasu
>            Assignee: Amareshwari Sri Ramadasu
>             Fix For: 0.16.0
>
>         Attachments: patch-2092.txt, patch-2092.txt, patch-2092.txt, 
> patch-2092.txt
>
>
> Pipes process does not die, when the java map task  dies.
> Pipes process does not die even when the cluster is down.

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

Reply via email to