[jira] [Commented] (FLUME-1946) Tolerate disk failure in FileChannel

["Hari Shreedharan (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FLUME-1946?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13600776#comment-13600776 ] 

Hari Shreedharan commented on FLUME-1946:
-----------------------------------------

It is incredibly hard to mirror files when there are multiple writers. Unless the order of writes to the files are exactly the same, the replay logic will end up in trouble (since we read puts/takes/commits in exactly the same order as the writes). Also, this is likely to make the channel code a lot more complex. That said, it is worth exploring various possibilities - definitely good to make the channel tolerant to disk failures.


                
> Tolerate disk failure in FileChannel
> ------------------------------------
>
>                 Key: FLUME-1946
>                 URL: https://issues.apache.org/jira/browse/FLUME-1946
>             Project: Flume
>          Issue Type: Bug
>          Components: File Channel
>            Reporter: Mike Percy
>
> It would be great to investigate changes to the FileChannel that enable it to tolerate failure of one or more disks in a group of data directories across disks without losing data. This involves implementing support for some form of mirroring at the software level, as well as appropriate recovery routines.

--
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