hadoop-yarn-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Junping Du (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (YARN-3160) Non-atomic operation on nodeUpdateQueue in RMNodeImpl
Date Mon, 09 Feb 2015 15:17:34 GMT

    [ https://issues.apache.org/jira/browse/YARN-3160?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14312342#comment-14312342
] 

Junping Du commented on YARN-3160:
----------------------------------

Hi [~chengbing.liu], thanks for reporting the issue here. However, the patch attached seems
only pull the head of queue (the oldest one) instead of pull all UpdatedContainerInfos in
nodeUpdateQueue which doesn't sounds right to me. Can you check the patch again?

> Non-atomic operation on nodeUpdateQueue in RMNodeImpl
> -----------------------------------------------------
>
>                 Key: YARN-3160
>                 URL: https://issues.apache.org/jira/browse/YARN-3160
>             Project: Hadoop YARN
>          Issue Type: Bug
>          Components: resourcemanager
>    Affects Versions: 2.7.0
>            Reporter: Chengbing Liu
>            Assignee: Chengbing Liu
>         Attachments: YARN-3160.patch
>
>
> {code:title=RMNodeImpl.java|borderStyle=solid}
> while(nodeUpdateQueue.peek() != null){
>   latestContainerInfoList.add(nodeUpdateQueue.poll());
> }
> {code}
> The above code brings potential risk of adding null value to {{latestContainerInfoList}}.
Since {{ConcurrentLinkedQueue}} implements a wait-free algorithm, we can directly poll the
queue, before checking whether the value is null.



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

Mime
View raw message