[
https://issues.apache.org/jira/browse/CASSANDRA-10007?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14662630#comment-14662630
]
Steve Wang commented on CASSANDRA-10007:
----------------------------------------
Seems to work fine in dtest, as Adam and I both found to be true.
The problem seems to be when the cluster is running on more than one node.
There's a direct correlation between the number of nodes and the number of rows
returned. In addition, there's an inverse relation between the fetch_size and
the number_of_rows returned. For example, with:
5 nodes, fetch_size = 3 i get values from "select* from test.test" oscillating
between 112 and 113, when the result should be 100.
In addition, the error only seems to occur when there isn't a "limit".
> Repeated rows in paged result
> -----------------------------
>
> Key: CASSANDRA-10007
> URL: https://issues.apache.org/jira/browse/CASSANDRA-10007
> Project: Cassandra
> Issue Type: Bug
> Components: Core
> Reporter: Adam Holmberg
> Assignee: Benjamin Lerer
> Labels: client-impacting
> Fix For: 3.x
>
> Attachments: paging-test.py
>
>
> We noticed an anomaly in paged results while testing against 3.0.0-alpha1. It
> seems that unbounded selects can return rows repeated at page boundaries.
> Furthermore, the number of repeated rows seems to dither in count across
> consecutive runs of the same query.
> Does not reproduce on 2.2.0 and earlier.
> I also noted that this behavior only manifests on multi-node clusters.
> The attached script shows this behavior when run against 3.0.0-alpha1.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
