Not sure if this is the problem but the default in 0.7* is to used framed transport, which means creating the TFramedTransport rather than TBufferedTransport. 

How are you connecting with python? Is it using framed transport?

Hope that helps. 

On 09 Nov, 2010,at 07:55 AM, David Replogle <> wrote:

I'm using Cassandra 0.7beta3 and it's running on localhost:9160 and Python works with it just fine. So, I go to run C++ against the system and I get:

TTransportException: No more data to read.

I did the thrift --gen cpp interface/cassandra.thrift in my 0.7beta3 folder
then included those files in the C++ code, and it builds fine, and runs.

It bombs out on the set_keyspace call... the code is short so I'll paste it here (not counting includes and using calls):


const string host("localhost");
const int port= 9160;

int main()
    try {
        boost::shared_ptr<TSocket> socket(new TSocket(host, port));
        boost::shared_ptr<TTransport> transport(new TBufferedTransport(socket));
        boost::shared_ptr<TProtocol> protocol(new TBinaryProtocol(transport));

        CassandraClient client(protocol);

        string version;
    } catch (apache::thrift::transport::TTransportException &tte) {
        cout << "TTransportException: " << tte.what() << endl;
    return 0;


It throws this same error with a series of other calls. This is probably (hopefully) something stupid I'm missing. Do I really need to run the C++ server component (which seems superfluous since it's not actually my Cassandra instance) as detailed here: http://wiki.apacheorg/thrift/ThriftUsageC%2B%2B

David Replogle | Senior Programmer Analyst

Catalyst SGC
44 Grandville Ave SW, Ste. 270, Grand Rapids, MI 49503
Office: 616.855.5522 x204 | Mobile: 616.293.2788