[OMPI users] Fault Tolerance

[Mohammad Huwaidi]

The following code is my trial to write a fault-tolerant application on 
OpenMPI; however, it still does not trap exceptions:

#include <mpi.h>
//#include <mpi++.h>
#include <iostream>
#include <unistd.h>
#include <cstdio>
#include <fstream>
#include <string>

#define BUFSIZE 100

using namespace std;
using namespace MPI;

static int nerr = 0;
static int iter = 0;

static ofstream fout;

void mpiErrorFunction( MPI::Comm &comm, int *code, ... )
{
    cout << "Encountered an MPI error No. " << ++nerr << endl << flush;
    fout << "Encountered an MPI error No. " << ++nerr << endl << flush;
    fout.flush();
}


void errhandler()
{
    cout << "Encountered an MPI error No. " << ++nerr << endl << flush;
    fout << "Encountered an MPI error No. " << ++nerr << endl << flush;
    fout.flush();
}


int main(int argc, char *argv[])
{
    cout << "Welcome to MPI error handling testing." << endl << flush;
    fout.flush();

    int  buf[BUFSIZE];
    char msg[BUFSIZE];
    MPI::Status status;
    char fileName[100];
    char buff[1000];

    cout << "Initializing MPI." << endl << flush;
    fout.flush();
    MPI::Init_thread(MPI::THREAD_MULTIPLE);
    //MPI::Init(argc, argv);
    cout << "MPI Initialized." << endl << flush;
    fout.flush();

    int myrank   = MPI::COMM_WORLD.Get_rank();
    int numprocs = MPI::COMM_WORLD.Get_size();

    sprintf(fileName,"MPItesting%i_%i.txt", myrank, numprocs);
    fout.open(fileName);

    sprintf(buff, "My task id is %d & No. of procs is %i\n", myrank, 
numprocs);
    cout << buff << flush;
    fout << buff << flush;
    fout.flush();

    if (myrank == 0) {
       //MPI::Errhandler errHandler =
       //    MPI::Comm::Create_errhandler (mpiErrorFunction);
       //MPI::COMM_WORLD.Set_errhandler(errHandler);
       //MPI::COMM_WORLD.Set_errhandler(MPI::ERRORS_RETURN);
       MPI::COMM_WORLD.Set_errhandler(MPI::ERRORS_THROW_EXCEPTIONS);
       while (nerr < 20 && iter++ < 100) {
          for (int i=1; i<numprocs; i++) {
              sprintf(msg,"%3i Hello there No. %i.", iter, i);
              sprintf(buff, "Message %i is being sent to %i\n", iter,  i);
              cout << buff << flush;
              fout << buff << flush;
              fout.flush();
              try {
                MPI::COMM_WORLD.Send(msg, BUFSIZE, CHAR, i, 99);
              }
              catch (MPI::Exception ex) {
                cout << "Could not send successfully to task No. " << i 
<< endl << flush;
                fout << "Could not send successfully to task No. " << i 
<< endl << flush;
                fout.flush();
              }
              sleep(1);
          }
       }
    }
    else {
         while (nerr < 20 && iter++ < 100) {
           try {
             MPI::COMM_WORLD.Recv(msg, BUFSIZE, CHAR, 0, 99);
             sprintf(buff,"%3i Receiving message No. %s\n", iter, msg);
              cout << buff << flush;
              fout << buff << flush;
              fout.flush();
           }
           catch (MPI::Exception ex) {
             cout << "Could not receive successfully "<< endl << flush;
             fout << "Could not receive successfully "<< endl << flush;
             fout.flush();
           }
           //sleep(myrank);
         }
    }

    cout << "Task no. " << myrank << " is existing now." << endl << flush;
    fout << "Task no. " << myrank << " is existing now." << endl << flush;
    fout.flush();
    fout.close();
    MPI::Finalize();                // MPI
    return 0;
}
--

Regards,
Mohammad Huwaidi

We can't resolve problems by using the same kind of thinking we used
when we created them.
                                                --Albert Einstein
<<attachment: mohammad.vcf>>