hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Colin Patrick McCabe (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-5323) Remove some deadcode in BlockManager
Date Tue, 08 Oct 2013 16:08:42 GMT

    [ https://issues.apache.org/jira/browse/HDFS-5323?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13789328#comment-13789328
] 

Colin Patrick McCabe commented on HDFS-5323:
--------------------------------------------

The only time {{BlocksMap#blocks}} is reassigned is when the {{BlocksMap}} is closed-- at
which time, {{blocks}} is set to null.  This would result in an NPE if anyone tried to do
anything with the map.  It should be fine, as well as more performant, to make it non-volatile.

> Remove some deadcode in BlockManager
> ------------------------------------
>
>                 Key: HDFS-5323
>                 URL: https://issues.apache.org/jira/browse/HDFS-5323
>             Project: Hadoop HDFS
>          Issue Type: Improvement
>          Components: namenode
>    Affects Versions: 2.3.0
>            Reporter: Colin Patrick McCabe
>            Assignee: Colin Patrick McCabe
>            Priority: Minor
>
> {{BlockManager#DEFAULT_MAP_LOAD_FACTOR}} is deadcode.  It no longer does *anything* since
blocks are now stored in a GSet whose size is fixed.
> {{BlocksMap#blocks}} does not need to be volatile.  Whenever it is accessed, it is accessed
under the {{FSNamesystem}} lock.  Furthermore, access to this data structure is not thread-safe.



--
This message was sent by Atlassian JIRA
(v6.1#6144)

Mime
View raw message