[
https://issues.apache.org/jira/browse/HBASE-720?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Izaak Rubin updated HBASE-720:
------------------------------
Attachment: hbase-720.patch
I've attached a patch (hbase-720.patch). It modifies the compaction process
such that compaction *does not* remove cells that have been marked as deleted
(now it only removes cells that have exceed the version number or ttl). I've
also updated TestCompaction to reflect these changes.
All HBase tests passed successfully, except for TestMasterAdmin and
TestRegionServerExit (which both failed before applying the patch as well).
> clean up inconsistencies around deletes
> ---------------------------------------
>
> Key: HBASE-720
> URL: https://issues.apache.org/jira/browse/HBASE-720
> Project: Hadoop HBase
> Issue Type: Bug
> Reporter: stack
> Assignee: Izaak Rubin
> Attachments: hbase-720.patch
>
>
> Compaction clears all cells behind any found 'delete' cell. This makes for
> an inconsistency in that before the compaction runs, you can do a get or scan
> on a timestamp older than the delete cell and get results. Confusing.
> Either preserve all cells for all time (or until > MAX_VERSIONS or TTL) or
> else make it so presence of a delete cell in the future stops client fetching
> anything older (expensive).
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
