[ 
https://issues.apache.org/jira/browse/CASSANDRA-7241?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14003413#comment-14003413
 ] 

Sylvain Lebresne commented on CASSANDRA-7241:
---------------------------------------------

If others run into that NPE in MiniDFSCluster.startDataNodes, the 
problem/solution is explained 
[here|http://stackoverflow.com/questions/17625938/hbase-minidfscluster-java-fails-in-certain-environments].
 With that fixed, I do get the same testCqlStorageCompositeKeyTable failure.

That said, when starting to look into what's going on, I added one simple log 
statement in ModificationStatement to log the applied mutation, and somehow 
that was enough to make the test pass some of the times (not every time, but 
often enough). So it's a timing issue, probably something that's not really 
related to CASSANDRA-5417 but was exposed by it because the timing changed. 
Where to go from there, I'm not sure. I have zero experience with Pig or Hadoop 
and our integration of it, and I have a hard time wrapping my head around the 
CqlStorage/AbstractCassandraStorage code tbh, so I'm not sure I'm the most 
qualified person to track this race down. But given that the pig test seems to 
be the only ones having problem of that sort, it suggests (though I can't be 
sure) that said race is more likely in either the pig code or the pig tests 
than anything else.

So I'm going to unassigned myself for now because I just don't have the time 
and experience to dig into the pig layers. If someone more familiar with those 
is able to reduce the failure to something Cassandra specific, I'll gladly take 
a look at that.

> Pig test fails on 2.1 branch
> ----------------------------
>
>                 Key: CASSANDRA-7241
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-7241
>             Project: Cassandra
>          Issue Type: Bug
>            Reporter: Alex Liu
>            Assignee: Sylvain Lebresne
>
> run ant pig-test on cassandra-2.1 branch. There are many tests failed. I 
> trace it a little and find out Pig test fails starts from 
> https://github.com/apache/cassandra/commit/362cc05352ec67e707e0ac790732e96a15e63f6b
> commit.
> It looks like storage changes break Pig tests.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Reply via email to