You are viewing a plain text version of this content. The canonical link for it is here.

Posted to commits@cassandra.apache.org by "Benedict (JIRA)" <ji...@apache.org> on 2014/02/05 11:42:08 UTC

[jira] [Commented] (CASSANDRA-6652) Stop CommitLogSegment.close() from unnecessarily calling sync() prior to cleaning the buffer

    [ https://issues.apache.org/jira/browse/CASSANDRA-6652?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13891980#comment-13891980 ] 

Benedict commented on CASSANDRA-6652:
-------------------------------------

Attaching a patch that solves this problem for 2.0.4, and also tries particularly hard to make sure the file is zeroed/unreadable by replay in the event that we fail to delete it.

> Stop CommitLogSegment.close() from unnecessarily calling sync() prior to cleaning the buffer
> --------------------------------------------------------------------------------------------
>
>                 Key: CASSANDRA-6652
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-6652
>             Project: Cassandra
>          Issue Type: Improvement
>            Reporter: Benedict
>            Priority: Minor
>         Attachments: tmp-2.0.patch
>
>
> What it says on the tin. Currently CLS.close() calls sync() before cleaning the buffer, which may result in unnecessary IO, but also in the case of a failure on the commit disk may prevent CLS from being discarded, or in the case of a faulty file may prevent any progress on reclaiming CLS.



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)