You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@cassandra.apache.org by "Aleksey Yeschenko (JIRA)" <ji...@apache.org> on 2015/09/19 14:42:04 UTC

[jira] [Commented] (CASSANDRA-10376) Fix upgrade_tests/paging_test.py:TestPagingData.static_columns_paging_test

    [ https://issues.apache.org/jira/browse/CASSANDRA-10376?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14877088#comment-14877088 ] 

Aleksey Yeschenko commented on CASSANDRA-10376:
-----------------------------------------------

To quote [~bdeggleston] again:
{quote}
Here is the commit that fixes the failure. This test is still failing, but this time it's because there's a NPE on the 2.1 node, although the test itself isn't failing any assertions.

Regarding the NPE, the cause appears to be the 3.0 node sending the normal cells for a partition in the first read, then only the static cells for the same partition in the subsequent reads, which causes AbstractQueryPager.firstNonStaticCell to return null.
{quote}

> Fix upgrade_tests/paging_test.py:TestPagingData.static_columns_paging_test
> --------------------------------------------------------------------------
>
>                 Key: CASSANDRA-10376
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-10376
>             Project: Cassandra
>          Issue Type: Sub-task
>            Reporter: Aleksey Yeschenko
>            Assignee: Blake Eggleston
>             Fix For: 3.0.x
>
>
> Follow-up to CASSANDRA-10354 to fix the related upgrade issue.
> To quote [~bdeggleston]:
> {quote}
> So the failure is caused by an edge case where a names filter is used in a paging query against a table that needs SinglePartitionNamesCommand instances converted to SinglePartitionSliceCommand instances in order to be converted to legacy read commands.
> If the previous read returned all requested clusterings and a number of rows equal to the page size, the subsequent read would have an empty clustering names filter. When an empty clustering names filter is converted to a slice filter, the slice filter is created with Slices.ALL.
> {quote}



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