Bug#1066735: mpich: fails to connect processes and report ranks with trivial mpi test

Samuel Thibault sthibault at debian.org
Tue Mar 26 17:38:22 GMT 2024


Hello,

Samuel Thibault, le ven. 15 mars 2024 10:31:54 +0100, a ecrit:
> Lucas Nussbaum, le mer. 13 mars 2024 15:56:40 +0100, a ecrit:
> > > [P0T0] Starting EZTrace (pid: 878489)...
> > > [P0T0] MPI mode selected
> > > This program requires 2 MPI processes, aborting...
> > > dir: mpi_ping_trace
> > > /bin/rm: cannot remove 'mpi_ping_trace': Directory not empty
> > > [P0T0] Stopping EZTrace (pid:878489)...
> > > [P0T0] Starting EZTrace (pid: 878488)...
> > > [P0T0] MPI mode selected
> > > This program requires 2 MPI processes, aborting...
> > > [P0T0] Stopping EZTrace (pid:878488)...
> > >      [OK] 
> 
> The test does run 2 processes. I tried this:
> 
> $ cat test.c
> #include <mpi.h>
> #include <stdio.h>
> int main(int argc, char *argv[]) {
> 	int rank, size;
> 	MPI_Init(&argc, &argv);
> 	MPI_Comm_rank(MPI_COMM_WORLD, &rank);
> 	MPI_Comm_size(MPI_COMM_WORLD, &size);
> 	printf("I'm %d/%d\n", rank, size);
> 	return 0;
> }
> 
> And it reports:
> 
> $ mpirun -np 2 ./test
> Authorization required, but no authorization protocol specified
> 
> Authorization required, but no authorization protocol specified
> 
> Authorization required, but no authorization protocol specified
> 
> Authorization required, but no authorization protocol specified
> 
> I'm 0/1
> I'm 0/1
> 
> and the same with a hosts file containing localhost twice.

I tried with disabling PMIX (commenting PMIX:=
--with-pmix=/usr/lib/$(DEB_HOST_MULTIARCH)/pmix2), and that fixed it.

Unless somebody complains, I will NMU that change, to get back mpich
working in unstable.

Samuel



More information about the debian-science-maintainers mailing list