[ 
https://issues.apache.org/jira/browse/OOZIE-2568?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15331536#comment-15331536
 ] 

WangMeng commented on OOZIE-2568:
---------------------------------

[~puru] [~satyaharish] [~shwethags] [~virag] [~devaraj]  ,anyone can review 
this simple patch ? thanks.

> SSH  action pretends to retry automaticly when it failed
> --------------------------------------------------------
>
>                 Key: OOZIE-2568
>                 URL: https://issues.apache.org/jira/browse/OOZIE-2568
>             Project: Oozie
>          Issue Type: Bug
>          Components: core
>    Affects Versions: 4.2.0
>            Reporter: WangMeng
>         Attachments: OOZIE-2568.01.patch
>
>
> There is a bug in automaticly retry of SSH action :
> For example:
> I have configed the following retry property :
> {code}
> <name>oozie.service.LiteWorkflowStoreService.user.retry.error.code.ext</name>
> <value>ALL</value>
> {code}
> And my SSH action is :
> {code}
>   <action name="ssh-afbb" retry-max="3" retry-interval="1">
>         <ssh xmlns="uri:oozie:ssh-action:0.1">
>             <host>wangmeng@XXXX</host>
>             <command>sh /data/wangmeng/hue_sh4.sh</command>
>             <capture-output/>
>         </ssh>
>         <ok to="End"/>
>         <error to="Kill"/>
>     </action>
> {code}
> When this action failed,it pretends to retry automaticly according to logs. 
> Such as :
> {code}
>  Start action [0000000-160612140701137-oozie-oozi-W@ssh-afbb] with user-retry 
> state : userRetryCount [1], userRetryMax [3], userRetryInterval [1]
> {code}
> However, it does not actually re run the above command "sh 
> /data/wangmeng/hue_sh4.sh".
> The reason is : if  the previous PID exists in XXXX.pid file of this SSH 
> action’s log dir , without checking this PID process is finished or not , SSH 
> action will not launch a new process to rerun command. And in my tests , I 
> find this  PID process have finished when Oozie rerun this action.



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

Reply via email to