db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Mamta A. Satoor (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (DERBY-6283) indexStat daemon processing tables over and over even when there are no changes in the tables in soft upgraded database.
Date Sun, 14 Jul 2013 04:18:49 GMT

    [ https://issues.apache.org/jira/browse/DERBY-6283?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13707946#comment-13707946
] 

Mamta A. Satoor commented on DERBY-6283:
----------------------------------------

Mike, I reviewed the patch and here is my understanding of the patch
1)We will pick up single-column unique indexes stats(this is the existing behavior and not
touched by this jira) for disposing 
   a)if we are dealing with a 10.9 or higher version db and 
   b)if Property.STORAGE_AUTO_INDEX_STATS_DEBUG_KEEP_DISPOSABLE_STATS is set to FALSE or not
set at all
2)we will dispose off all the stats which are not attached to any indexes for the given conglomerate(this
is the new behavior with this jira) irrespective of what happens with single-column unique
index stats

Just two little nit-picks on the patch and an existing comment
1)Did you mean to write more in the comment below? It looks unfinished
            //  This loop looks for statistic entries to delete.  It deletes
            //  those entries that don't have a matching conglomerate in the
2)Also, does this existing comment still make sense? The above loop will include single column
unique index if we are dealing with 10.9 and higher db and if we are not being asked to keep
the stats for single column unique index through Property.STORAGE_AUTO_INDEX_STATS_DEBUG_KEEP_DISPOSABLE_STATS
flag.
        // Note that the algorithm would drop valid statistics entries if
        // working on a subset of the table conglomerates/indexes.
        // The above loop has populated "cds" with only existing indexes that
        // are not single column unique.

                
> indexStat daemon processing tables over and over even when there are no changes in the
tables in soft upgraded database.
> ------------------------------------------------------------------------------------------------------------------------
>
>                 Key: DERBY-6283
>                 URL: https://issues.apache.org/jira/browse/DERBY-6283
>             Project: Derby
>          Issue Type: Bug
>          Components: Store
>    Affects Versions: 10.8.3.1, 10.9.2.2, 10.10.1.3, 10.11.0.0
>            Reporter: Mike Matrigali
>            Assignee: Mike Matrigali
>            Priority: Blocker
>              Labels: derby_triage10_11
>         Attachments: derby-6283.diff
>
>
> The fix for DERBY-5680 currently requires hard upgrade.  A soft upgrade database with
an orphaned stat row, that can be caused by something like DERBY-5681, will cause DERBY to
spin eating up 100% of a cpu and possibly all the disk bandwith of the disk where the database
is located.
> We should implement a fix that can be applied to soft upgraded databases if at all possible.
> In soft upgraded databases we can no use the work around of dropping the statistics as
it is not available
> to soft upgraded db's with version < 10.9.  The only current workaround is to disable
background stats
> completely.  Dropping and recreating the suspect table may also work.

--
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