cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sam Tunnicliffe (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-9062) Investigate failing collection indexing dtests
Date Thu, 09 Apr 2015 17:13:13 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-9062?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14487678#comment-14487678
] 

Sam Tunnicliffe commented on CASSANDRA-9062:
--------------------------------------------

This looks to be a regression caused by CASSANDRA-8793 (git bisect agrees). The issue is triggered
by the fact that the keys in the 2i table are UUIDs. The RowPosition passed to BigTableReader#getPosition
uses a heap buffer, with default byte order. When keys are then read back from the summary
in IndexSummary#binarySearch the hollow byte buffer has native ordering so the 2 buffers don't
compare as expected, which ultimately causes the sstable to be skipped. 

[~benedict] I confirmed that forcing the byte order of the hollow buffer to BigEndian in IndexSummary#binarySearch
fixes this specific case but I'm sure there must be a better solution, any suggestions?

> Investigate failing collection indexing dtests
> ----------------------------------------------
>
>                 Key: CASSANDRA-9062
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-9062
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Tests
>            Reporter: Tyler Hobbs
>            Assignee: Sam Tunnicliffe
>             Fix For: 3.0
>
>
> There are frequent failures with the dtests related to indexing collections ({{secondary_indexes_test.py:TestSecondaryIndexesOnCollections}}).
> I tried to look into the reason for failure.  The test does seem to be correct, because
it can occasionally pass.  It seems like the failures are due to indexing lagging behind the
inserts, resulting in a partial set of results when the index is queried shortly afterwards
(getting ~20k matching rows instead of the expected 50k).



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

Mime
View raw message