cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Benjamin Lerer (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-11467) Paging loses rows in certain conditions
Date Fri, 01 Apr 2016 09:32:25 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-11467?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15221452#comment-15221452
] 

Benjamin Lerer commented on CASSANDRA-11467:
--------------------------------------------

The problem is impacting tables with compact storage and no clustering keys (the partitioner
does not impact the behaviour). 
When a range query is performed, if the page is not the first one, an extra row is requested
to the storage engine and removed by the {{AbstractQueryPager}}. 
In the case where the first columns is not live the extra row is not fully removed by {{AbstractQueryPager}}.
The row will be ignored when the result set is built but will be taken into account by {{AbstractQueryPager}}
as the last row returned. By consequence, the row will not be returned in the next page.

[~rhatch] The problem does not seems related to CASSANDRA-11195

> Paging loses rows in certain conditions
> ---------------------------------------
>
>                 Key: CASSANDRA-11467
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-11467
>             Project: Cassandra
>          Issue Type: Bug
>            Reporter: Ian McMahon
>            Assignee: Benjamin Lerer
>         Attachments: pagination_test.go
>
>
> The bug occurs under the following conditions:
>   - RandomPartitioner
>   - a compact storage CF
>   - querying across rows
>   - a tombstone in the first column of a row on the pagesize boundary
>   - you need to be querying at least 2*pagesize + 1 records
> Attached is a go program using gocql which reproduces the bug fairly minimally.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message