You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@cassandra.apache.org by "amorton (JIRA)" <ji...@apache.org> on 2012/12/20 09:09:12 UTC

[jira] [Commented] (CASSANDRA-5079) Compaction deletes ExpiringColumns in Secondary Indexes

    [ https://issues.apache.org/jira/browse/CASSANDRA-5079?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13536855#comment-13536855 ] 

amorton commented on CASSANDRA-5079:
------------------------------------

Yup, was a fun few hours.

bq. (Setting affects-version to when 2I were added. Please correct if this was a more recent regression.)
Yes it has always been there. 

Will need to patch 0.8 onwards. Do we still release 0.7?
                
> Compaction deletes ExpiringColumns in Secondary Indexes
> -------------------------------------------------------
>
>                 Key: CASSANDRA-5079
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-5079
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Core
>    Affects Versions: 0.7.0
>            Reporter: amorton
>            Assignee: amorton
>             Fix For: 1.2.0 rc2
>
>         Attachments: 5079.txt
>
>
> From this discussion http://www.mail-archive.com/user@cassandra.apache.org/msg26599.html
> CompactionManager.getDefaultGcBefore() set's the gc_before to be Integer.MAX_VALUE. 
> In the example all entries in the secondary index have a TTL. In PreCompactedRow.removeDeletedAndOldShards() the CF is determined to have irrelevant data, the call to CFS.removeDeleted() results in the ExpiringColumns being removed and the row is treated as empty. CompactionTask.execute() exits at the {{if (!nni.hasNext())}} test, so the sstables are marked as compacted and soon deleted. 
> In the example the localDeletionTime was Thu, 21 Mar 2013 08:25:22 GMT and should not have been purged. 
> In the example when the first compaction on the secondary index runs the on disk data for the index is deleted. The logs show a compaction starting and no associated "Compacted to" message from that compaction thread. 
> The impact is incorrect results from secondary indexes queries.

--
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