Tao Yang commented on YARN-8127:

Thanks [~cheersyang]. Some additional details: This problem is most often 
happened in async-scheduling mode, multiple async scheduling threads may 
allocate for the same node which has reserved container and generate duplicated 
allocate-from-reserved proposals for the same reserved container. These 
duplicated proposals can be successfully applied in commit phase so that 
resources of node/queue will be decreased several times for a same container. 

We can add a check of reserved state for allocate-from-reserved proposals in 
commit phase to solve this problem.

Attached a patch with UT for review. 

> Resource leak when async scheduling is enabled
> ----------------------------------------------
>                 Key: YARN-8127
>                 URL: https://issues.apache.org/jira/browse/YARN-8127
>             Project: Hadoop YARN
>          Issue Type: Bug
>            Reporter: Weiwei Yang
>            Assignee: Tao Yang
>            Priority: Critical
> Brief steps to reproduce
>  # Enable async scheduling, 5 threads
>  # Submit a lot of jobs trying to exhaust cluster resource
>  # After a while, observed NM allocated resource is more than resource 
> requested by allocated containers
> Looks like the commit phase is not sync handling reserved containers, causing 
> some proposal incorrectly accepted, subsequently resource was deducted 
> multiple times for a container.

This message was sent by Atlassian JIRA

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