hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Kihwal Lee (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-4972) [branch-0.23] permission check and operation are done in a separate lock for getBlockLocations()
Date Mon, 15 Jul 2013 20:32:49 GMT

    [ https://issues.apache.org/jira/browse/HDFS-4972?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13708942#comment-13708942
] 

Kihwal Lee commented on HDFS-4972:
----------------------------------

Thanks for the review, Daryn. I've committed this to branch-0.23.
                
> [branch-0.23] permission check and operation are done in a separate lock for getBlockLocations()
> ------------------------------------------------------------------------------------------------
>
>                 Key: HDFS-4972
>                 URL: https://issues.apache.org/jira/browse/HDFS-4972
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: namenode
>    Affects Versions: 0.23.8
>            Reporter: Kihwal Lee
>            Assignee: Kihwal Lee
>         Attachments: HDFS-4972-branch-0.23.patch
>
>
> For getBlockLocations() call, the read lock is acquired when doing permission check.
But unlike other namenode methods, this is outside of the lock of the actual operation. So
it ends up acquiring and releasing the lock twice.  This has two implications.
> - permissions can change in between the locks
> - the lock fairness will penalize getBlockLocations().
> This was fixed in trunk and branch-2 as a part of HDFS-4679, but not in branch-0.23.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message