[
https://issues.apache.org/jira/browse/CASSANDRA-9071?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14570683#comment-14570683
]
Fabien Rousseau commented on CASSANDRA-9071:
--------------------------------------------
Benedict, thanks for your patch.
I also provided a patch a few days ago (see attached file
cassandra-2.1-9071.txt, maybe I should have added a comment to make it more
explicit :) )
This is a similar approach, but mine only keep the DecoratedKey (there is no
need to keep the whole previous CF, and can consume more memory if the CF has
many columns)
Concerning the unit test, it needs to at least write first a full buffer then
write the second one (the first buffer never has an empty partition). Moreover,
to increase chances of triggering this false assertion, it’s better to have the
decorated key bigger than the value (and only have one column per CF). (This
increases the fact that, when “sync” is called, no more columns are written in
the empty CF)
Don’t hesitate to have a look at the unit test I submitted in the patch (I
tested it without (test failed) and with my patch (test passed) )
> CQLSSTableWriter gives java.lang.AssertionError: Empty partition
> ----------------------------------------------------------------
>
> Key: CASSANDRA-9071
> URL: https://issues.apache.org/jira/browse/CASSANDRA-9071
> Project: Cassandra
> Issue Type: Bug
> Components: Tools
> Environment: java 7 / 8
> cassandra 2.1.3 snapshot build locally with last commit
> https://github.com/apache/cassandra/commit/6ee4b0989d9a3ae3e704918622024fa57fdf63e7
> macos Yosemite 10.10.2
> Reporter: Ajit Joglekar
> Assignee: Fabien Rousseau
> Fix For: 2.1.x
>
> Attachments: EmailWriter.java, Screen Shot 2015-04-15 at 11.14.40
> PM.png, cassandra-2.1-9071.txt, data.csv.gz
>
>
> I am always getting the following error:
> Exception in thread "main" java.lang.AssertionError: Empty partition
> at
> org.apache.cassandra.io.sstable.SSTableSimpleUnsortedWriter$DiskWriter.run(SSTableSimpleUnsortedWriter.java:228)
> It happens at a certain point that seems to be repeatable. Only issue is I am
> converting 400 million records into multiple SSTables and creating small test
> is a challenge
> Last comment from Benedict looks relevant here
> https://issues.apache.org/jira/browse/CASSANDRA-8619
> Is there a work around, quick fix, fix that I can try out locally?
> Thanks,
> -Ajit
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
