[MPICH] Aborting: connection to smpd rejected

Goncalo Borges goncalo at lip.pt
Thu Apr 20 06:04:25 CDT 2006


Hi there,
I'm trying to use a tight integration of MPICH with SGE. According to the 
SGE staff how-to I should use the smpd mechanism for that. In my local cluster, 
everything works correctly using:

mpiexec -ssh -nopm -n $NSLOTS -machinefile $TMPDIR/machines $HOME/$FILE_EXE

The problem is that my cluster is integrated in a global grid.
Users which try to run mpi jobs though the grid (in my cluster) get the 
following error:

Aborting: connection to smpd rejected
Aborting: connection to smpd rejected
Aborting: connection to smpd rejected
Aborting: connection to smpd rejected

which I don't understand where it is coming from.
Could you explain what could be the reason of this error, keeping in mind 
that the stuff works for local users but not for the grid environment?

Thanks in advance
Cheers
Goncalo




More information about the mpich-discuss mailing list