db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Mike Matrigali (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (DERBY-6328) Improve Store error messages and repair tools for corrupt index conglomerates.
Date Fri, 30 Aug 2013 18:02:52 GMT

    [ https://issues.apache.org/jira/browse/DERBY-6328?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13754959#comment-13754959
] 

Mike Matrigali commented on DERBY-6328:
---------------------------------------

i agree it would good for these errors to at least provide a conglomerate number, and maybe
for a nested language level error to provide the table/index name.  Especially with write
cache enabled corrupted db's it is hard for store to know what it might see on a page.  

For instance in error case I think the page it is trying to read in just has never been formatted,
store likely requested a formated version of the page be forced to disk but it was not.  Part
of the page is the format id telling what object the page is, and if we can't identify
that object it is a serious error that is completely unexpected by store.  
                
> Improve Store error messages and repair tools for corrupt index conglomerates.
> ------------------------------------------------------------------------------
>
>                 Key: DERBY-6328
>                 URL: https://issues.apache.org/jira/browse/DERBY-6328
>             Project: Derby
>          Issue Type: Improvement
>          Components: SQL, Store
>    Affects Versions: 10.11.0.0
>            Reporter: Rick Hillegas
>
> If file system write-caching is enabled and the machine crashes, Derby conglomerates
can be corrupted. By default, write-caching is usually enabled. For performance reasons, applications
often leave it on despite our warning that this can lead to data corruption. Even if an application
vendor recommends the disabling of the write-cache, the end user may still choose to leave
it on. To reduce the tech-support burden for these applications, we could do the following:
> o Improve Store error messages for errors caused by corrupt conglomerates. If the Store
knows that a conglomerate is corrupt, the error message ought to be able to indicate which
conglomerate needs repair. XBM0U is an example of an error message which could carry this
extra information.
> o We could provide a system procedure which takes an index conglomerate number as an
argument and then drops and re-creates that conglomerate. This may result in a cascading series
of conglomerate recreations if foreign keys are involved. This could be cheaper than compressing
all of the tables in a big database after one conglomerate becomes corrupt.
> This enhancement request was suggested by a discussion on DERBY-5221.

--
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

Mime
View raw message