cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Stu Hood (JIRA)" <j...@apache.org>
Subject [jira] Commented: (CASSANDRA-1748) Flush before repair
Date Wed, 15 Dec 2010 01:48:01 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-1748?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12971520#action_12971520
] 

Stu Hood commented on CASSANDRA-1748:
-------------------------------------

> would switching validation to use range scans be a better solution?
We don't have a way to send a memtable to another node, so if we included it in what we validate,
it might not actually end up being sent to the other node unless it was flushed at some point.
So I don't think doing this would buy us much, and it is a significantly larger change. The
goal is really just to repair at a "point in time", and flushing before repair gives us that,
imo.

> Flush before repair
> -------------------
>
>                 Key: CASSANDRA-1748
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-1748
>             Project: Cassandra
>          Issue Type: Bug
>    Affects Versions: 0.7.0 rc 2
>            Reporter: Stu Hood
>            Assignee: Tyler Hobbs
>             Fix For: 0.7.1
>
>         Attachments: 1748-trunk.txt, 1748.txt
>
>
> We don't currently flush before beginning a validation compaction, meaning that depending
on the state of the memtables, we might end up with content on disk that is as different as
a single memtable can make it (potentially, very different).

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message