You are viewing a plain text version of this content. The canonical link for it is here.
Posted to yarn-issues@hadoop.apache.org by "Junping Du (JIRA)" <ji...@apache.org> on 2015/12/30 16:41:49 UTC

[jira] [Updated] (YARN-1382) Remove unusableRMNodesConcurrentSet (never used) in NodeListManager to get rid of memory leak

     [ https://issues.apache.org/jira/browse/YARN-1382?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Junping Du updated YARN-1382:
-----------------------------
    Summary: Remove unusableRMNodesConcurrentSet (never used) in NodeListManager to get rid of memory leak  (was: NodeListManager has a memory leak, unusableRMNodesConcurrentSet is never purged)

> Remove unusableRMNodesConcurrentSet (never used) in NodeListManager to get rid of memory leak
> ---------------------------------------------------------------------------------------------
>
>                 Key: YARN-1382
>                 URL: https://issues.apache.org/jira/browse/YARN-1382
>             Project: Hadoop YARN
>          Issue Type: Bug
>          Components: resourcemanager
>    Affects Versions: 2.2.0, 2.7.1, 2.6.2
>            Reporter: Alejandro Abdelnur
>            Assignee: Rohith Sharma K S
>         Attachments: 0001-YARN-1382.patch, 0002-YARN-1382.patch, 0003-YARN-1382.patch
>
>
> If a node is in the unusable nodes set (unusableRMNodesConcurrentSet) and never comes back, the node will be there forever.
> While the leak is not big, it gets aggravated if the NM addresses are configured with ephemeral ports as when the nodes come back they come back as new.
> Some related details in YARN-1343



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