Welcome, Guest
Username: Password: Remember me

TOPIC: cluster directory problem

cluster directory problem 14 years 3 weeks ago #818

  • asturner
  • asturner's Avatar
Hi,

I need to run my simulation with a cluster of 4 computers.
Right now, i'm working on 2, in order to find how that works but i've a little problem.

I got 2 hosts : canal_dell2 and canal_dell5

The mpi_telemac.conf contains this :


# Nombre de processeurs :
2
#
# Pour chaque host :
#
# nom_du_host nombre_de_processeurs_sur_ce_host
#
canal_dell2 1
canal_dell5 1


The systel.ini file contains :

#- réseau de PCs
RUN_MPI="mpiexec -file mpirun.txt"

The other lines are commented ( with # )


On both computers, smpd is running. Both users are registered ( mpiexec -validate -> SUCCESS ).

The files for the simulation are on both computers in the same directory ( C:\cas2000\ ), and both hard drives C: are shared on the network.

But when I run the simulation (from the host canal_dell2) here is the problem i get :




That is to say, the firewall is disabled.


Do you have any suggestion to solve this problem ?


Thanks,


Alexandre
The administrator has disabled public write access.

Re:cluster directory problem 14 years 3 weeks ago #819

  • c.coulet
  • c.coulet's Avatar
  • OFFLINE
  • Moderator
  • Posts: 3722
  • Thank you received: 1031
Hi
Try with another directory!
Drive C exist on the both machine so you create confusion
The files should be located on only one directory which is share between the computers

On both computer, map a directory P: on your working directory
try to execute telemac on the mapped drive

Regards
Christophe
The administrator has disabled public write access.

Re:cluster directory problem 14 years 3 weeks ago #820

  • asturner
  • asturner's Avatar
I'm trying right now on 2 computers with Windows 7 (at home)

I put all the simulation's files in "E:\test\", and I shared E:\ with all computers.

I've the simple error :

MPI launcher : mpiexec -file mpirun.txt
abort: Unable to connect to Alexandre-PC
Duration of job : 3 seconds (0:0:3 ) (system=0 sec)


I think that i don't really get it when you say : "map a directory P: ". You just mean that I've to share a special directory in which all my simulation files are contained and which exists only on one computer ?

Regards,


Alexandre
The administrator has disabled public write access.

Re:cluster directory problem 14 years 3 weeks ago #821

  • c.coulet
  • c.coulet's Avatar
  • OFFLINE
  • Moderator
  • Posts: 3722
  • Thank you received: 1031
Hi

P: is an example. I think your test is correct this time. The most important point for Telemac in parallel is that all processors should find all files (which are located in the tmp directory created during the launch of the program) at the same location (E:\test\ in your case)

Your problem comes from MPI.

Regards
Christophe
The administrator has disabled public write access.
Moderators: borisb

The open TELEMAC-MASCARET template for Joomla!2.5, the HTML 4 version.