Hi all,

we  solved the problem ... 

first we updated the Cassandra for 1.2.3 version and after another exception was launched "No hosts   to borrow from" and we discovered that the command "ConnectionPoolConfigurationImpl(...).setConnectTimeout(-1)" was the cause ...

and we put .setConnectTimeout(2000) and we increased the other values from  Astyanax pool and our App worked ... 

basically I think our initial problem was Amazon latency that was to high, so we changed our pool configuration and the things worked fine ...

thank all for the help !

Andre


2013/4/17 aaron morton <aaron@thelastpickle.com>
If you Hadoop task supplying both a start and finish key for the slice ? You probably only want the start. 

Provide the full call stack and the code in your hadoop task. 

Cheers
 
-----------------
Aaron Morton
Freelance Cassandra Consultant
New Zealand

@aaronmorton

On 18/04/2013, at 1:34 AM, "Hiller, Dean" <Dean.Hiller@nrel.gov> wrote:

What's the stack trace you see?  At the time, I was thinking column scan not row scan as perhaps your code or priam's code was doing a column slice within a row set and the columns are sorted by Integer while priam is passing in UTF8 or vice-versa.  Ie. Do we know if this is a column sorting issue or a row one?

Dean

From: Andre Tavares <andre271@gmail.com<mailto:andre271@gmail.com>>
Reply-To: "user@cassandra.apache.org<mailto:user@cassandra.apache.org>" <user@cassandra.apache.org<mailto:user@cassandra.apache.org>>
Date: Wednesday, April 17, 2013 7:09 AM
To: "user@cassandra.apache.org<mailto:user@cassandra.apache.org>" <user@cassandra.apache.org<mailto:user@cassandra.apache.org>>
Subject: Re: InvalidRequestException: Start key's token sorts after end token

Dean,

sorry,  but I saw your comments on Stackoverflow (http://stackoverflow.com/questions/16041727/operationtimeoutexception-cassandra-cluster-aws-emr ) just after I sent this message ...

and I think you may be right about the sort method,  but Priam sets  Cassandra partitioner with "RandomPartitioner", and maybe the correct could be "Murmur3Partitioner" when we use Hadoop (I am not sure too) ... if that is true I got a problem because I can't change the partitioner with Priam (I think it only works with RandomPartitioner) ...

Andre

2013/4/17 Hiller, Dean <Dean.Hiller@nrel.gov<mailto:Dean.Hiller@nrel.gov>>
I literally jut replied to your stackoverflow comment then saw this email.  I need the whole stack trace.  My guess is the ColFamily is configured for one sort method where map/reduce is using another or something when querying but that's just a guess.

Dean

From: Andre Tavares <andre271@gmail.com<mailto:andre271@gmail.com><mailto:andre271@gmail.com<mailto:andre271@gmail.com>>>
Reply-To: "user@cassandra.apache.org<mailto:user@cassandra.apache.org><mailto:user@cassandra.apache.org<mailto:user@cassandra.apache.org>>" <user@cassandra.apache.org<mailto:user@cassandra.apache.org><mailto:user@cassandra.apache.org<mailto:user@cassandra.apache.org>>>
Date: Wednesday, April 17, 2013 6:47 AM
To: "user@cassandra.apache.org<mailto:user@cassandra.apache.org><mailto:user@cassandra.apache.org<mailto:user@cassandra.apache.org>>" <user@cassandra.apache.org<mailto:user@cassandra.apache.org><mailto:user@cassandra.apache.org<mailto:user@cassandra.apache.org>>>
Subject: InvalidRequestException: Start key's token sorts after end token

know what exactly this message means a