[
https://issues.apache.org/jira/browse/CASSANDRA-6893?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13946747#comment-13946747
]
Sylvain Lebresne commented on CASSANDRA-6893:
---------------------------------------------
Actually, we kind of have a dtest already:
[paxos_test|https://github.com/riptano/cassandra-dtest/blob/master/paxos_tests.py].
It's doing something very similar to the java example above. However, it's
running on a relatively modest number of threads so that it isn't reproducing
this bug on my box with the currently hardcoded setting. But if I change said
setting to match the ones of the test above, i.e. 1000 threads with 1 iteration
each, then it does reproduce.
So anyway, I modified said test so it runs with both the old parameters, and
the new ones, the ones that reproduce this issue (let me note that I did run
into CASSANDRA-6923 while reproducing, but that's a red herring). So anyway, we
do have a dtest now.
> Unintended update with conditional statement
> --------------------------------------------
>
> Key: CASSANDRA-6893
> URL: https://issues.apache.org/jira/browse/CASSANDRA-6893
> Project: Cassandra
> Issue Type: Bug
> Environment: Ubuntu Precise 64bit / Cassandra 2.0.6
> Reporter: Suguru Namura
> Assignee: Sylvain Lebresne
> Fix For: 2.0.7
>
> Attachments: 6893.txt, ConcurrentCASUpdate.java
>
>
> After updated to 2.0.6, I have encountered the strange behavior of
> conditional updates.
> When I executed CQL like UPDATE test SET value = ? WHERE id = ? IF value = ?
> in concurrent, sometimes cassandra returns true even if value is not
> satisfied the condition.
> I have attached the program which reproduce this issue. The program works
> fine in cassandra 2.0.5. But it seems that resets values while execution in
> 2.0.6.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
