hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Clint Morgan (JIRA)" <j...@apache.org>
Subject [jira] Updated: (HBASE-476) RegexpRowFilter behaves incorectly when there are multiple store files
Date Wed, 05 Mar 2008 00:49:40 GMT

     [ https://issues.apache.org/jira/browse/HBASE-476?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

Clint Morgan updated HBASE-476:

    Attachment: hbase-476-test2.patch

This is a cleaner way to provoke the issue.

- Write enough to trigger a flush
- Let all flushes complete
- Write a bit more to the memcache
- Open up a scanner with a RegExpRowFilter
- Scanner fails to provide any results if some columns are in memcache, and some are in hstores.

Hopefully someone familiar with the Scanner implementation can take a look at it.

> RegexpRowFilter behaves incorectly when there are multiple store files
> ----------------------------------------------------------------------
>                 Key: HBASE-476
>                 URL: https://issues.apache.org/jira/browse/HBASE-476
>             Project: Hadoop HBase
>          Issue Type: Bug
>          Components: filters
>            Reporter: Clint Morgan
>         Attachments: hbase-476-test.log, hbase-476-test.patch, hbase-476-test2.patch
> I noticed that after running some table Map/Reduces, then using a
> RegExpRowFilter to scan through the table,  the scanner misses
> rows when its columns are in different stores.
> This (rather convoluted) unit test provokes the behavior.
> - Set memcache flush size small to trigger multiple stores
> - put in 10 row with 2 columns. Each row has the same value for col1 (which the RowFilter
wants to match)
> - Scan with and without the filter to be sure that we get all the rows with each
> - Run an identity table M/R 10 times to fill up the memcache and trigger flush.
> - Scan again. This time the filter does not pickup anything.
> Attaching the log from this run as well.

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message