Hi Marcos,
Yes, I have a problem with spawning to a “worker” host (on localhost, works).
There are just two machines: “master” and “worker”. I’m using Windows 10 in
both with same Cygwin and packages. Pasted below some details.
Thanks for your help. Regards,
Martín
----
Running:
mpirun -np 1 -hostfile ./hostfile ./spawner.exe 8
hostfile:
master slots=5
worker slots=5
Error:
At least one pair of MPI processes are unable to reach each other for
MPI communications. This means that no Open MPI device has indicated
that it can be used to communicate between these processes. This is
an error; Open MPI requires that all MPI processes be able to reach
each other. This error can sometimes be the result of forgetting to
specify the "self" BTL.
Process 1 ([[31598,1],0]) is on host: DESKTOP-C0G4680
Process 2 ([[31598,2],2]) is on host: worker
BTLs attempted: self tcp
Your MPI job is now going to abort; sorry.
--------------------------------------------------------------------------
[DESKTOP-C0G4680:02828] [[31598,1],0] ORTE_ERROR_LOG: Unreachable in file
/pub/devel/openmpi/v4.0/openmpi-4.0.5-1.x86_64/src/openmpi-4.0.5/ompi/dpm/dpm.c
at line 493
[DESKTOP-C0G4680:02828] *** An error occurred in MPI_Comm_spawn
[DESKTOP-C0G4680:02828] *** reported by process [2070806529,0]
[DESKTOP-C0G4680:02828] *** on communicator MPI_COMM_SELF
[DESKTOP-C0G4680:02828] *** MPI_ERR_INTERN: internal error
[DESKTOP-C0G4680:02828] *** MPI_ERRORS_ARE_FATAL (processes in this
communicator will now abort,
[DESKTOP-C0G4680:02828] *** and potentially your MPI job)
USER_SSH@DESKTOP-C0G4680 ~
$ [WinDev2012Eval:00120] [[31598,2],2] ORTE_ERROR_LOG: Unreachable in file
/pub/devel/openmpi/v4.0/openmpi-4.0.5-1.x86_64/src/openmpi-4.0.5/ompi/dpm/dpm.c
at line 493
[WinDev2012Eval:00121] [[31598,2],3] ORTE_ERROR_LOG: Unreachable in file
/pub/devel/openmpi/v4.0/openmpi-4.0.5-1.x86_64/src/openmpi-4.0.5/ompi/dpm/dpm.c
at line 493
--------------------------------------------------------------------------
It looks like MPI_INIT failed for some reason; your parallel process is
likely to abort. There are many reasons that a parallel process can
fail during MPI_INIT; some of which are due to configuration or environment
problems. This failure appears to be an internal failure; here's some
additional information (which may only be relevant to an Open MPI
developer):
ompi_dpm_dyn_init() failed
--> Returned "Unreachable" (-12) instead of "Success" (0)
--------------------------------------------------------------------------
[WinDev2012Eval:00121] *** An error occurred in MPI_Init
[WinDev2012Eval:00121] *** reported by process
[15289389101093879810,12884901891]
[WinDev2012Eval:00121] *** on a NULL communicator
[WinDev2012Eval:00121] *** Unknown error
[WinDev2012Eval:00121] *** MPI_ERRORS_ARE_FATAL (processes in this communicator
will now abort,
[WinDev2012Eval:00121] *** and potentially your MPI job)
[DESKTOP-C0G4680:02831] 2 more processes have sent help message
help-mca-bml-r2.txt / unreachable proc
[DESKTOP-C0G4680:02831] Set MCA parameter "orte_base_help_aggregate" to 0 to
see all help / error messages
[DESKTOP-C0G4680:02831] 1 more process has sent help message
help-mpi-runtime.txt / mpi_init:startup:internal-failure
[DESKTOP-C0G4680:02831] 1 more process has sent help message
help-mpi-errors.txt / mpi_errors_are_fatal unknown handle
Script spawner:
#include "mpi.h"
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
int main(int argc, char ** argv){
int processesToRun;
MPI_Comm intercomm;
MPI_Info info;
if(argc < 2 ){
printf("Processes number needed!\n");
return 0;
}
processesToRun = atoi(argv[1]);
MPI_Init( NULL, NULL );
printf("Spawning from parent:...\n");
MPI_Comm_spawn( "./spawned.exe", MPI_ARGV_NULL, processesToRun,
MPI_INFO_NULL, 0, MPI_COMM_SELF, &intercomm, MPI_ERRCODES_IGNORE);
MPI_Finalize();
return 0;
}
Script spawned:
#include "mpi.h"
#include <stdio.h>
#include <stdlib.h>
int main(int argc, char ** argv){
int hostName_len,rank, size;
MPI_Comm parentcomm;
char hostName[200];
MPI_Init( NULL, NULL );
MPI_Comm_get_parent( &parentcomm );
MPI_Get_processor_name(hostName, &hostName_len);
MPI_Comm_rank(MPI_COMM_WORLD, &rank);
MPI_Comm_size(MPI_COMM_WORLD, &size);
if (parentcomm != MPI_COMM_NULL) {
printf("I'm the spawned h: %s r/s: %i/%i\n", hostName, rank, size );
}
MPI_Finalize();
return 0;
}
From: Marco Atzeri via users<mailto:users@lists.open-mpi.org>
Sent: miércoles, 3 de febrero de 2021 17:58
To: users@lists.open-mpi.org<mailto:users@lists.open-mpi.org>
Cc: Marco Atzeri<mailto:marco.atz...@gmail.com>
Subject: Re: [OMPI users] OMPI 4.1 in Cygwin packages?
On 03.02.2021 21:35, Martín Morales via users wrote:
> Hello,
>
> I would like to know if any OMPI 4.1.* is going to be available in the
> Cygwin packages.
>
> Thanks and regards,
>
> Martín
>
Hi Martin,
anything in it that is abolutely needed short term ?
Any problem with current 4.0.5 package ?
Usually it is very time consuming the build
and I am busy with other cygwin stuff
Regards
Marco
