cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Dominic Williams (JIRA)" <j...@apache.org>
Subject [jira] Created: (CASSANDRA-1423) Cassandra's internal state broke by getting column slices. Error org.apache.thrift.TApplicationException: Internal error processing get_slice afterwards.
Date Mon, 23 Aug 2010 21:38:16 GMT
Cassandra's internal state broke by getting column slices. Error org.apache.thrift.TApplicationException:
Internal error processing get_slice afterwards.
---------------------------------------------------------------------------------------------------------------------------------------------------------

                 Key: CASSANDRA-1423
                 URL: https://issues.apache.org/jira/browse/CASSANDRA-1423
             Project: Cassandra
          Issue Type: Bug
          Components: Core
    Affects Versions: 0.7 beta 1
         Environment: Ubuntu 9.04
            Reporter: Dominic Williams
            Priority: Critical
             Fix For: 0.7 beta 2


The attached program (with data that it can import) causes large column slices to be requested
from Cassandra. 

The program itself uploads school address data to a Cassyndex full text index, and then allows
you to search that. The program simulates someone typing a search into an active search box,
which shows you the matches for the current term as you type. Thus when you enter a search
term such as "cherwell school oxford" actually it performs the searches "c", "ch", "che",
"cher", "cherw" etc

You can configure the delay between the "keystrokes". If your delay allows the searches to
complete sequentially, you are ok. But if you have a short delay, and searches are created
in parallel, pretty quickly this error will arise - "org.apache.thrift.TApplicationException:
Internal error processing get_slice".

Once this has occurred all future attempts and getting slices of columns will return the same
error, and your'e only option is to restart Cassandra.

This looks like some kind of concurrency edge condition bug caused by requesting sufficiently
large intersecting slices in parallel. It may be in other versions too.

I've been testing on 0.7 B1 using an RP cluster. 

The attached maven project should pull down the scale7 libraries but if interested you can
find the sources at http://github.com/s7

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message