[jira] [Resolved] (OOZIE-2075) hadoop jobs doesnot get killed when coordinator action get killed

["Rohini Palaniswamy (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/OOZIE-2075?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Rohini Palaniswamy resolved OOZIE-2075.
---------------------------------------
    Resolution: Duplicate

  Came across this issue recently myself and checked with our hadoop core team. In Hadoop 2, kill in YARN is SIGTERM followed 250ms later by SIGKILL. So most of the time shutdown hook would have very little time to run via a kill. Will address this partly in OOZIE-1025 by killing jobs using application tags and in another jira killing oozie child jobs from the action-data.seq file when launcher has managed to write the file successfully during a kill.  In the case I saw the file was created successfully by the launcher. So that should help releases using pre Hadoop 2.4 to kill child jobs at least in scenarios where the file was managed to be written. 

> hadoop jobs doesnot get killed when coordinator action get killed
> -----------------------------------------------------------------
>
>                 Key: OOZIE-2075
>                 URL: https://issues.apache.org/jira/browse/OOZIE-2075
>             Project: Oozie
>          Issue Type: Bug
>            Reporter: Praveen Adlakha
>            Assignee: Jaydeep Vishwakarma
>         Attachments: stdout.txt
>
>
> Hi ,
> Whenever the coord action or workflow gets killed only the job that has error gets killed while the other hadoop jobs keeps on running and only the workflow status gets changed to killed.
> Please add a feature to correct this behaviour.
> Thanks
> Praveen



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)