[ 
https://issues.apache.org/jira/browse/YARN-9662?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Amithsha updated YARN-9662:
---------------------------
    Description: 
Preemption on node labels is not working when the utilization is 100%.

Example
 adhocp0,adhocp1,adhocp3 Queues mapped to nodelabels of label_adhoc_nm.
 With a share of 60,30,10 as actual capacity and 100 as maximum capacity for 
all.
 When a jobA on adhocp3 consumes 100% of its maximum capacity and a jobB 
submitted on adhocp0 no containers running on adhocp3  queue got preempted.
  
 This is already reported by another user
 https://issues.apache.org/jira/browse/YARN-7685

Note :

Jobs with more than actual capacity and less than the maximum capacity are able 
to preempt the containers.

  was:
Preemption on node labels is not working at 100% utilisation.

Example
adhocp0,adhocp1,adhocp3 mapped to nodelabels of label_adhoc_nm.
With a share of 60,30,10 as actual capacity and 100 as maximum capacity for all.
When a jobA on adhocp3 consumes 100% of its maximum capacity and a jobB 
submitted on adhocp0 no containers running on adhocp3 got preempted.
 
This is already reported by another user
https://issues.apache.org/jira/browse/YARN-7685


> Preemption not working on NodeLabels
> ------------------------------------
>
>                 Key: YARN-9662
>                 URL: https://issues.apache.org/jira/browse/YARN-9662
>             Project: Hadoop YARN
>          Issue Type: Bug
>    Affects Versions: 2.9.0
>            Reporter: Amithsha
>            Priority: Major
>
> Preemption on node labels is not working when the utilization is 100%.
> Example
>  adhocp0,adhocp1,adhocp3 Queues mapped to nodelabels of label_adhoc_nm.
>  With a share of 60,30,10 as actual capacity and 100 as maximum capacity for 
> all.
>  When a jobA on adhocp3 consumes 100% of its maximum capacity and a jobB 
> submitted on adhocp0 no containers running on adhocp3  queue got preempted.
>   
>  This is already reported by another user
>  https://issues.apache.org/jira/browse/YARN-7685
> Note :
> Jobs with more than actual capacity and less than the maximum capacity are 
> able to preempt the containers.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
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