hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Aaron T. Myers (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-4777) File creation with overwrite flag set to true results in logSync holding namesystem lock
Date Tue, 30 Apr 2013 16:54:16 GMT

    [ https://issues.apache.org/jira/browse/HDFS-4777?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13645717#comment-13645717
] 

Aaron T. Myers commented on HDFS-4777:
--------------------------------------

Thanks a lot for working on this, Suresh.

>From a quick look it seems like most (all?) of these test failures fall into three categories:

# Holding the lock during sync because of the startFile/deleteInternal issue you initially
identified in this JIRA.
# Holding the lock when calling FSNS#logUpdateMasterKey when updating token secret keys.
# Holding the lock when calling FSNS#logSyncAll when entering safe mode.

The first two definitely seem like they should be fixed. The third I'm not so sure can/should
be fixed, since we're not really concerned about NN operation throughput when entering safemode,
and it does seem to be a good idea to call logSyncAll when entering SM.

Certainly feel free to make sub-tasks if you want, though it doesn't seem crazy to me to do
this in one JIRA. I don't think it will actually require changing all that many places.
                
> File creation with overwrite flag set to true results in logSync holding namesystem lock
> ----------------------------------------------------------------------------------------
>
>                 Key: HDFS-4777
>                 URL: https://issues.apache.org/jira/browse/HDFS-4777
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: namenode
>    Affects Versions: 0.23.0, 2.0.0-alpha
>            Reporter: Suresh Srinivas
>            Assignee: Suresh Srinivas
>            Priority: Blocker
>         Attachments: HDFS-4777.patch
>
>
> FSNamesystem#startFileInternal calls delete. Delete method releases the write lock, making
parts of startFileInternal code unintentionally executed without write lock being held.

--
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