[ 
https://issues.apache.org/jira/browse/YARN-5829?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15945821#comment-15945821
 ] 

ASF GitHub Bot commented on YARN-5829:
--------------------------------------

Github user kambatla commented on a diff in the pull request:

    https://github.com/apache/hadoop/pull/201#discussion_r108518561
  
    --- Diff: 
hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler/fair/FSSchedulerNode.java
 ---
    @@ -110,16 +140,56 @@ synchronized FSAppAttempt getReservedAppSchedulable() 
{
       }
     
       /**
    +   * List reserved resources after preemption and assign them to the
    +   * appropriate applications in a FIFO order.
    +   * @return if any resources were allocated
    +   */
    +  @VisibleForTesting
    +  synchronized LinkedHashMap<FSAppAttempt, Resource> getPreemptionList() {
    +    cleanupPreemptionList();
    +    return new LinkedHashMap<>(resourcesPreemptedForApp);
    +  }
    +
    +  /**
    +   * Remove apps that have their preemption requests fulfilled
    +   */
    +  private synchronized void cleanupPreemptionList() {
    +    Iterator<FSAppAttempt> iterator =
    +        resourcesPreemptedForApp.keySet().iterator();
    +    while (iterator.hasNext()) {
    +      FSAppAttempt app = iterator.next();
    +      if (app.isStopped() || !app.isStarved()) {
    +        // App does not need more resources
    +        Resources.subtractFrom(totalResourcesPreempted,
    +            resourcesPreemptedForApp.get(app));
    +        appIdToAppMap.remove(app.getApplicationAttemptId());
    +        iterator.remove();
    +      }
    +    }
    +  }
    +
    +  /**
        * Mark {@code containers} as being considered for preemption so they are
        * not considered again. A call to this requires a corresponding call to
    -   * {@link #removeContainerForPreemption} to ensure we do not mark a
    +   * {@code releaseContainer} for preemption to ensure we do not mark a
    --- End diff --
    
    Should we drop "for preemption" altogether? 


> FS preemption should reserve a node before considering containers on it for 
> preemption
> --------------------------------------------------------------------------------------
>
>                 Key: YARN-5829
>                 URL: https://issues.apache.org/jira/browse/YARN-5829
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: fairscheduler
>            Reporter: Karthik Kambatla
>            Assignee: Miklos Szegedi
>
> FS preemption evaluates nodes for preemption, and subsequently preempts 
> identified containers. If this node is not reserved for a specific 
> application, any other application could be allocated resources on this node. 
> Reserving the node for the starved application before preempting containers 
> would help avoid this.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

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