[
https://issues.apache.org/jira/browse/OOZIE-2668?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15523866#comment-15523866
]
Purshotam Shah commented on OOZIE-2668:
---------------------------------------
Ignore my previous comment. Wf can be killed even if it's in suspended/failed
state.
You need to add coordAction.incrementAndGetPending(), in if condition. We
should only increment pending when a command is queued.
> Status update and recovery problems when coord action and its children not in
> sync
> ----------------------------------------------------------------------------------
>
> Key: OOZIE-2668
> URL: https://issues.apache.org/jira/browse/OOZIE-2668
> Project: Oozie
> Issue Type: Bug
> Reporter: Satish Subhashrao Saley
> Assignee: Satish Subhashrao Saley
> Attachments: OOZIE-2668-1.patch, OOZIE-2668-2.patch
>
>
> In cases where workflow is already in terminal status (except failed) but the
> coord action is not yet updated and still running, following will happen if a
> kill command is issued on the coord job:
> Kill on Coordjob will make the kill on coordaction pending until the children
> are also killed. However, as the wf in terminal state (except failed), the wf
> will not be killed and preverifycondition will fail. The wf doesn't update
> its parent and hence the coordaction kill will still be pending. Two
> problems: Status transit service will not resolve the state of this coord job
> as some the actions are still pending Recovery service will try to recover
> this killed coord action and keep on reissuing the kill command.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
