[ 
https://issues.apache.org/jira/browse/YARN-10109?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17029546#comment-17029546
 ] 

Sunil G commented on YARN-10109:
--------------------------------

[~prabhujoseph] Thanks for the patch.

I am kind of +0 for having UNDEFINED state for Queue. I presume the Queue will 
land on to UNDEFINED state during some race condition scenarios. Post this, how 
a queue can be recovered? What user need to do at this point of time ? This is 
my assumption now. Please check below scenario for more details.

*My scenario is as follows:* User is NOT invoking validate api, But calling 
mutation api directly (may be wrong payload). If the payload config alone is 
updated with UNDEFINED state for some queue due to issues, and later you are 
discarding this config, then i think it's fine. I would like ensure that actual 
Queue in memory should never land on to UNDEFINED state in any case. 

> Allow stop and convert from leaf to parent queue in a single Mutation API call
> ------------------------------------------------------------------------------
>
>                 Key: YARN-10109
>                 URL: https://issues.apache.org/jira/browse/YARN-10109
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: capacity scheduler
>    Affects Versions: 3.3.0
>            Reporter: Prabhu Joseph
>            Assignee: Prabhu Joseph
>            Priority: Major
>         Attachments: YARN-10109-001.patch, YARN-10109-002.patch, 
> YARN-10109-003.patch
>
>
> SchedulerConf Mutation API does not Allow Stop and Adding queue under an 
> existing Leaf Queue in a single call. 
> *Repro:*
>  
> {code:java}
> Capacity-Scheduler.xml: 
> yarn.scheduler.capacity.root.queues = default
> yarn.scheduler.capacity.root.default.capacity = 100 
> cat abc.xml 
> <sched-conf>
>       <add-queue>
>           <queue-name>root.default.v1</queue-name>
>           <params>
>             <entry>
>               <key>capacity</key>
>               <value>100</value>
>             </entry>
>           </params>
>         </add-queue>
>         <update-queue>
>           <queue-name>root.default</queue-name>
>           <params>
>             <entry>
>               <key>state</key>
>               <value>STOPPED</value>
>             </entry>
>           </params>
>         </update-queue>
>       </sched-conf>  
> [yarn@pjoseph-1 tmp]$ curl --negotiate -u : -X PUT -d @add.xml -H 
> "Content-type: application/xml" 
> 'http://<RM_HOST>:8088/ws/v1/cluster/scheduler-conf?user.name=yarn'
> Failed to re-init queues : Can not convert the leaf queue: root.default to 
> parent queue since it is not yet in stopped state. Current State : RUNNING
>  {code}



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org

Reply via email to