[
https://issues.apache.org/jira/browse/CASSANDRA-884?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12873785#action_12873785
]
AJ Slater commented on CASSANDRA-884:
-------------------------------------
Please re-open or duplicate.
Testing with 0.6-trunk today:
Reading with CL > ONE returns multiple copies of the same column per key
consistent with the replicas queried before return. i.e, for RC=3, a QUORUM
read yields 2 copies and an ALL read returns 3.
This is with pycassa get_range() which is using get_range_slice()
I see the same behavior with 0.6.1 and 0.6.2 debs
If my experience is not unique, anyone using get_range_slice is now deluged
with duplicate data.
> get_range_slice returns multiple copies of each row for ConsistencyLevel > ONE
> ------------------------------------------------------------------------------
>
> Key: CASSANDRA-884
> URL: https://issues.apache.org/jira/browse/CASSANDRA-884
> Project: Cassandra
> Issue Type: Bug
> Components: Core
> Affects Versions: 0.5
> Environment: 4-cluster Gentoo Linux 2.6.18 with a ReplicationFactor
> of 2
> Reporter: Omer van der Horst Jansen
> Assignee: Omer van der Horst Jansen
> Fix For: 0.6
>
> Attachments: 0001-RangeSliceResponseResolver.patch, 884-v2.txt,
> TestApp2.java
>
>
> I've noticed that both 0.5.1 and 0.6b2 return multiple identical copies of
> the data stored in my keyspace whenever I make a call to get_range_slice or
> get_range_slices using
> ConsistencyLevel.QUORUM and ReplicationFactor is greater than one.
> So with ReplicationFactor set to 2 for my application's KeySpace I get double
> the number of KeySlices that I expect to get. When using ConsistencyLevel.ONE
> I get only one KeySlice for each row.
> I've seen this happen with Cassandra 0.5.1 and with 0.6 beta 2. The behavior
> on 0.6 beta 2 is exhibited with both get_range_slice and get_range_slices.
> The attached Java program demonstrates the issue for 0.6 beta 2. The program
> writes a series of single-column rows into the Standard1 table, and then uses
> get_range_slice to receive a list of all row. The returned number of rows is
> consistently twice the number of rows written to the database. I wipe out the
> database completely before running the test.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
