cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sylvain Lebresne (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-2576) Rewrite into new file post streaming
Date Fri, 29 Apr 2011 17:08:03 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-2576?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13027079#comment-13027079
] 

Sylvain Lebresne commented on CASSANDRA-2576:
---------------------------------------------

Looks good, but I would want to remove the addition of the close() function to SSTableWriter
(because it calls closeAndOpenReader). I fear people would be tempted to call close in a final
block, which could remove the temporary marker, even though some error would have left the
sstable corrupted. And as far as I can tell, this is a gratuitous addition, that has nothing
to do with that patch.

Writing a new file was in the first versions of the counter patch and was changed for the
space efficiency it provides (or at least I don't think there was another reason). Nevertheless
I do agree that it is cleaner that way, so let's hope the temporary increased space needed
won't bite anyone (since we delete the streamed file as soon as the new file is created, it's
probably fine).

> Rewrite into new file post streaming
> ------------------------------------
>
>                 Key: CASSANDRA-2576
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-2576
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Core
>            Reporter: Stu Hood
>            Assignee: Stu Hood
>             Fix For: 1.0
>
>         Attachments: 0001-Don-t-depend-on-a-byte-for-byte-match-for-SSTable-rebu.txt,
0002-Rebuild-into-a-new-file-to-minimize-magic.txt
>
>
> Commutative/counter column families use a separate path to rebuild sstables post streaming,
and that path currently rewrites the data within the streamed file. While this is great for
space efficiency, it means a duplicated code path for writing sstables, which makes it more
difficult to make changes like #674.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message