You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@cassandra.apache.org by "Jonathan Ellis (JIRA)" <ji...@apache.org> on 2009/05/15 19:23:45 UTC

[jira] Commented: (CASSANDRA-182) CommitLog.add doesn't really force to disk

    [ https://issues.apache.org/jira/browse/CASSANDRA-182?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12709914#action_12709914 ] 

Jonathan Ellis commented on CASSANDRA-182:
------------------------------------------

The more I look at the commitlog code the more I'm inclined to go with my initial reaction, that we shouldn't mess with this for 0.3... :-|

The FastSync option looks like A&P's attempt to deal with this.  It does call force() on close, for instance.  (Is that really all we need for durability?)  Having two options for syncing, one of which is commented to have issues and the other is virtually entirely untested, doesn't really seem like an improvement to me.

If we did go with "let's add a third option" then IMO we should be doing the force in CommitLog not the writer.  updateHeader for instance does a looping write w/ a single writer; i assume only one sync for the entire method is needed, not one sync per write.

For 0.3 I suggest emailing A&P off-list and see if they can clarify why fastsync has problems purging log files.  If it is something simple to fix great.  Otherwise let's note to "use fastsync for durable writes, but this has the drawback of not purging log files.  fixing this is a priority for 0.4"

> CommitLog.add doesn't really force to disk
> ------------------------------------------
>
>                 Key: CASSANDRA-182
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-182
>             Project: Cassandra
>          Issue Type: Bug
>    Affects Versions: trunk, 0.3
>            Reporter: Sandeep Tata
>            Assignee: Sandeep Tata
>             Fix For: trunk, 0.3
>
>         Attachments: CASSANDRA-182.patch
>
>
> CommitLog.add does't really force writes to disk. This could result in acked writes being lost.

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