[mpich-discuss] Problem with running mpiexec on more than oneWindows XP machines

[Jayesh Krishna]

Hi,
 I would recommend the following,
 
# Use the "-machinefile" option of mpiexec (mpiexec -n 2 -machinefile
mf.txt cpi.exe) to run your MPI program across multiple machines. See the
windows developer's guide for more information (available at
http://www.mcs.anl.gov/research/projects/mpich2/documentation/index.php?s=
docs).
# Register the username/password using "mpiexec -register" command instead
of using the textfile to store them. After you register your
username/password, mpiexec will automatically use it to launch your MPI
job.
# Use ipaddresses instead of host names.
# Make sure that you turn off any firewalls (windows, third-party) running
on the machines.
# Make sure that you can run a simple non-MPI program on the remote host.
        On machine1, mpiexec -n 1 -host machine2_ipaddress hostname
        On machine2, mpiexec -n 1 -host machine1_ipaddress hostname
 
 Let us know the results.
 
Regards,
Jayesh


  _____  

From: mpich-discuss-bounces at mcs.anl.gov
[mailto:mpich-discuss-bounces at mcs.anl.gov] On Behalf Of Sole Eko
Sent: Wednesday, February 18, 2009 1:01 PM
To: mpich-discuss at mcs.anl.gov
Subject: [mpich-discuss] Problem with running mpiexec on more than
oneWindows XP machines


Hi, 

I am new in MPI and MPICH in particular.

I firstly installed mpich2 on two machines in same lan, both machines have
same (admin) account name/password.
so I created file pwd.exe to store username and password.

than, I had set available hosts for smpd on both of them like this: 

smpd -sethosts ipmachine1 ipmachine2 , where I put real ip addresses of
machines.
___________________________
I tried to run example cpi.exe on machine1:

mpiexec.exe -pwdfile pwd.txt cpi.exe  - and it worked

than similar from machine1 started cpi.exe on machine2:

mpiexec.exe -pwdfile pwd.txt -host ipmachine2 cpi.exe - this also worked,
no problem
__________________________

but, when I tried to run on both machines, first like this:
mpiexec.exe -pwdfile pwd.txt -n 2 cpi.exe - it doesen't work, it start cpi
on machine1 and asks for number of intervals and then nothing is gonig
on.. just cousor in new line blinking..
and, when I cancel job with ctrl+c - i got this:

job aborted:
rank: node: exit code[: error message]
0: ipmachine1 : 123: mpiexec aborting job
1: ipmachine2 : 123
Terminate batch job (Y/N)?

so, after all, there is something happening on machine2, but I don't
understand what is going on..

I also tried with -machinefile option..  -  no results

I compiled cpi.c file and tired non-interactive version, same I got..

Can anybody help me?

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20090218/7a69e292/attachment.htm>