[OMPI users] Is it possible for independently started processes to connect via ompi-server? (1.8.1)

[James Barron]

Hello. I have read a couple of posts on Stackoverflow that seem to imply
that independently started processes should be able to communicate via
ompi-server, but I have not been able to accomplish this. Is this actually
possible? Here is an outline of what I have done and the error I am
receiving. I'm using version 1.8.1.

> ompi-server --no-daemonize -r +
78512128.0;tcp://192.168.1.219:50443

> mpirun -np 1 --hostfile ~/mpi-hosts --ompi-server "78512128.0;tcp://
192.168.1.219:50443" -v port_server
Master of rank 0 of 1 on kurenai
Master port is 488308736.0;tcp://192.168.1.219:60050
+488308737.0;tcp://192.168.1.219:48993:300

port_server code outline:

    char port[MPI::MAX_PORT_NAME];
    MPI::Intercomm intercomm;
    MPI::Info info;
    int buffer[3];

    MPI::Open_port(MPI::INFO_NULL, port);

    printf("Master port is %s\n", port);

    info = MPI::Info::Create();

    info.Set("ompi_global_scope", "true");

    MPI::Publish_name("test_service", info, port);

    intercomm = MPI::COMM_WORLD.Accept(port, MPI::INFO_NULL, 0);

    buffer[0] = 1;
    buffer[1] = 2;
    buffer[2] = 3;

    intercomm.Send(buffer, 3, MPI::INT, 0, PARENT_TAG);

    ...

> mpirun -np 1 --hostfile ~/mpi-hosts --ompi-server "78512128.0;tcp://
192.168.1.219:50443" -v port_client
barronj@192.168.1.219's password:
Slave of rank 0 of 1 on athena
Slave found test_service on port, 488308736.0;tcp://192.168.1.219:60050
+488308737.0;tcp://192.168.1.219:48993:300
[athena:08054] [[27027,0],0]-[[7451,0],0] mca_oob_tcp_peer_send_handler:
invalid connection state (6) on socket 19

client code outline:

    char port[MPI::MAX_PORT_NAME];
    MPI::Intercomm intercomm;
    int buffer[3];

    MPI::Lookup_name("test_service", MPI::INFO_NULL, port);

    printf("Slave found test_service on port, %s\n", port);

    intercomm = MPI::COMM_WORLD.Connect(port, MPI::INFO_NULL, 0);

    intercomm.Recv(buffer, 3, MPI::INT, 0, PARENT_TAG);

    ...

>From testing, it looks like it is happening within Connect. What am I doing
incorrectly?

Thank you,

James Barron