Do you use the identical version of openmpi on both machines, configured in the same way?




On Sep 30, 2015, at 18:29, Andre Vieira <apvieira@if.usp.br> wrote:

Hello,

I would be glad if ALPS users could help me with a problem I've been experiencing. I have ALPS 2.2.b3 installed in an iMac (i5 mid 2011) and an i7 desktop running Ubuntu 15.04 64 bits, with openmpi (1.6.5) also present. I have no problem running ALPS executables in either computer, nor running basic programs using openmpi (for instance `mpirun -host <host> hostname`) from either machine to the other. I can also successfully run ALPS executables in either machine using openmpi locally. 

However, when I try to run an ALPS executable (e.g. loop) via openmpi using both machines simultaneously, the jobs are initiated, keep running indefinitely, but no output is produced, not even to the screen. Using mpirun with the -v flag gives no additional information. All I get is the following:

dirac-2:qmc apvieira$ mpirun -v -np 7 loop --mpi heisloop20.in.xml --Tmin 5
ALPS/looper version 3.2b12-20100128 (2010/01/28)
  multi-cluster quantum Monte Carlo algorithms for spin systems
  copyright (c) 1997-2010 by Synge Todo <wistaria@comp-phys.org>

using ALPS/parapack scheduler
  a Monte Carlo scheduler for multiple-level parallelization
  copyright (c) 1997-2013 by Synge Todo <wistaria@comp-phys.org>

based on the ALPS libraries version 2.2.b3
  available from http://alps.comp-phys.org/
  copyright (c) 1994-2013 by the ALPS collaboration.
  Consult the web page for license details.
  For details see the publication: 
  B. Bauer et al., J. Stat. Mech. (2011) P05001.


Did anyone have this kind of problem before?

Thanks,
Andre Vieira