Diff between aster-mpihosts and mpi-hostfile

JMB365 · 2011-10-16 18:25:34

Hello,

In the path /opt/aster/etc/codeaster/ there are two files aster-mpihosts and mpi-hostfile. What is the difference between the two and what should the typical contents be?

To me it seems that one of them is un-necessary. Can somebody kindly explain or refute my thinking? Thanks

Regards, JMB

Last edited by JMB365 (2011-10-16 23:25:22)

Thomas DE SOZA · 2011-10-16 19:00:04

Hi,

None of these files are mandatory. It all depends on what you have entered in your asrun config file. Here's an excerpt of mine regarding MPI configuration :

As I do not reference %(mpi_hostfile)s the host file is not needed and on my computer it does not exist (running a MPI version only for developement and debut purposes I do not run on more than one compute node).

# ----- MPI commands and parameters
# mpirun
#    available arguments are : mpi_hostfile, mpi_nbnoeud, mpi_nbcpu
#    (use Python string formatting style)
mpirun_cmd : mpirun --tag-output -np %(mpi_nbcpu)s %(program)s

# file which contains list of hosts (REQUIRED even if it is not used in mpirun_cmd)
mpi_hostfile : /opt/aster/etc/codeaster/aster-mpihosts

# command called to initialize MPI environment (for example for mpich2)
#mpi_ini : mpdboot --totalnum=%(mpi_nbnoeud)s --file=%(mpi_hostfile)s ; sleep 10

# command called to close the MPI session (for example for mpich2)
#mpi_end : mpdallexit

# shell command to get processor id
# LAM/MPI : echo $LAMRANK
# OpenMPI (1.2) : echo $OMPI_MCA_ns_nds_vpid
# OpenMPI (1.3) : echo $OMPI_MCA_orte_ess_vpid
# OpenMPI (1.34 : echo $OMPI_COMM_WORLD_RANK
# Mpich2  : echo $PMI_RANK
mpi_get_procid_cmd : echo $OMPI_COMM_WORLD_RANK

TdS

JMB365 · 2011-10-17 00:09:22

Hello TdS,

Thank you for replying to my queries even on a Sunday afternoon! I really appreciate such good help and so timely too...

For the most part I have been running only one compute node and am working my way into a multi-core multi-node cluster. So far I have only tested mpiexec between two PCs (a single core & the other is a quad-core). Running through the "Troubleshooting Tips" of the MPICH2 User manual enabled me to get mpdboot working between the two and the ability to run a test program like calculating Pi (Pi.c) in this 2 member ring.

However I am stuck at the point where if I submit a CodeAster job (the test problem mumps01a.comm) via ASTK using ncpus=1; mpi_nbcpu=2 & mpi_nbnoeud=2, the terminal window pops up and the job is stuck forever at:

 Code_Aster run
<INFO> Command line 1 :
<INFO> ./asteru_mpi Python/Execution/E_SUPERV.py -eficas_path ./Python -commandes fort.1 -rep none  -num_job 8551-ubuntu32 -mode interactif -rep_outils /opt/aster/outils -rep_mat /opt/aster/NEW11.0_mpi/materiau -rep_dex /opt/aster/NEW11.0_mpi/datg -suivi_batch -memjeveux 64.0 -tpmax 5400

Prior to submitting the CodeAster job I started mpd using:

mpdboot -n 2

Because without it the ASTK job crashes saying mpd is not running. So I am trying to muddle my way around trying to get CodeAster to actually run in this 2 PC test cluster. Any hints you can throw my way, I would really be grateful for. Thanks a bunch for the help you have given over the years!

Regards, JMB

Last edited by JMB365 (2011-10-17 00:10:31)

JMB365 · 2011-10-17 00:13:41

Hey TdS,

I just read you message more carefully and I think I see the answer. I think if I un-comment the mpi_ini & mpi_end command I might get somewhere! Hope I am right...

Regards, JMB

#1 2011-10-16 18:25:34

Diff between aster-mpihosts and mpi-hostfile

#2 2011-10-16 19:00:04

Re: Diff between aster-mpihosts and mpi-hostfile

#3 2011-10-17 00:09:22

Re: Diff between aster-mpihosts and mpi-hostfile

#4 2011-10-17 00:13:41

Re: Diff between aster-mpihosts and mpi-hostfile

Board footer