[MPICH] MPICH2 on WinXP-Pro cluster

[David Ashton]

Peter,

So did you try running your job like you mentioned in your previous email?

"C:\PROGRA~1\MPICH2\bin>mpiexec -pwdfile Options.txt -exitcodes -hosts 2
familycomputer raethserver cpi"

Did this fail after installing the latest version but it worked with an
older version?

-David Ashton

-----Original Message-----
From: PETER RAETH [mailto:PeterRaeth at email.uophx.edu] 
Sent: Sunday, August 07, 2005 10:09 PM
To: David Ashton
Subject: Re: [MPICH] MPICH2 on WinXP-Pro cluster

Thanks for working with me on this David.

Downloaded the latest version of MPICH2 and installed it on both machines, 
after uninstalling the older version.

During the run attempt, raethserver, rejected my credentials.

Further examination showed that there is no domain configuration, just a 
workgroup. Tried it without specifying the workgroup and it will not work 
that way. Will talk to our networking people in the morning since I am not 
sure how to configure the system for domains.


Best,

Peter.

Peter Raeth, Ph.D., Faculty
University of Phoenix (Online Campus)
PeterRaeth at email.uophx.edu
http://www.geocities.com/peter_raeth/


----- Original Message ----- 
From: "David Ashton" <ashton at mcs.anl.gov>
To: "'PETER RAETH'" <PeterRaeth at email.uophx.edu>
Cc: <mpich-discuss at mcs.anl.gov>
Sent: Saturday, August 06, 2005 1:12 PM
Subject: RE: [MPICH] MPICH2 on WinXP-Pro cluster


> Peter,
>
> It appears from the error output that you are not running the latest 
> release
> of MPICH2.  Can you download and install 1.0.2?
> http://www.mcs.anl.gov/mpi/mpich2
>
> Also, the username has to be the same on both machines.  Do you have a
> domain called MSHOME or are your machines in a workgroup?  If it isn't a
> domain then I would leave the MSHOME\ part off of the username.
>
> -David Ashton
>
> -----Original Message-----
> From: PETER RAETH [mailto:PeterRaeth at email.uophx.edu]
> Sent: Saturday, August 06, 2005 11:53 AM
> To: David Ashton
> Subject: Re: [MPICH] MPICH2 on WinXP-Pro cluster
>
> David,
>
> Made sure both machines were fully updated with the latest WinXP-Pro
> operating system.
>
> Went to both machines and turned off the firewall on each. Verified that
> smpd.exe is running on both.
>
> Tried the cpi example again with the following result. familycomputer is 
> the
>
> main cluster node. raethserver is the worker node connected via a 
> cross-over
>
> cable.
>
> C:\Program Files\MPICH2\bin>ping raethserver
>
> Pinging raethserver [192.168.1.25] with 32 bytes of data:
>
> Reply from 192.168.1.25: bytes=32 time<1ms TTL=128
> Reply from 192.168.1.25: bytes=32 time<1ms TTL=128
> Reply from 192.168.1.25: bytes=32 time<1ms TTL=128
> Reply from 192.168.1.25: bytes=32 time<1ms TTL=128
>
> Ping statistics for 192.168.1.25:
>    Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
> Approximate round trip times in milli-seconds:
>    Minimum = 0ms, Maximum = 0ms, Average = 0ms
>
> C:\Program Files\MPICH2\bin>mpiexec -hosts 1 familycomputer cpi
> User credentials needed to launch processes:
> account (domain\user): MSHOME\Peter Raeth
> password:
> Enter the number of intervals: (0 quits) 100
> pi is approximately 3.1416009869231254, Error is 0.0000083333333323
> wall clock time = 0.000042
> Enter the number of intervals: (0 quits) 200
> pi is approximately 3.1415947369231252, Error is 0.0000020833333321
> wall clock time = 0.000043
> Enter the number of intervals: (0 quits) 0
>
> C:\Program Files\MPICH2\bin>mpiexec -hosts 2 raethserver familycomputer 
> cpi
> Aborting: 2 unable to connect to raethserver
>
>
> Many thanks for whatever insights you can offer.
>
>
> Best,
>
> Peter.
>
> Peter Raeth, Ph.D., Faculty
> University of Phoenix (Online Campus)
> PeterRaeth at email.uophx.edu
> http://www.geocities.com/peter_raeth/
>
>
> ----- Original Message ----- 
> From: "David Ashton" <ashton at mcs.anl.gov>
> To: "'Peter Raeth (FAC)'" <PeterRaeth at email.uophx.edu>;
> <mpich-discuss at mcs.anl.gov>
> Sent: Friday, August 05, 2005 12:15 PM
> Subject: RE: [MPICH] MPICH2 on WinXP-Pro cluster
>
>
>> Peter,
>>
>> What is the error output?  Can you send us that?
>>
>> If mpiexec can resolve the ip address of the other machine but cannot
>> connect to it then it is probably a firewall issue.  The simple answer is
>> that you need to disable the firewall to run MPI applications.
>>
>> The longer answer is that if you are using the Windows firewall then you
>> can
>> configure it to allow the smpd.exe process not to be blocked.  Then you
>> need
>> to allow exceptions for each MPI executable that you intend to run.  This
>> way you can still have the firewall running but it won't block ports used
>> by
>> mpiexec or your application.
>>
>> -David Ashton
>>
>> -----Original Message-----
>> From: owner-mpich-discuss at mcs.anl.gov
>> [mailto:owner-mpich-discuss at mcs.anl.gov] On Behalf Of Peter Raeth (FAC)
>> Sent: Friday, August 05, 2005 11:13 AM
>> To: 'mpich-discuss at mcs.anl.gov'
>> Subject: [MPICH] MPICH2 on WinXP-Pro cluster
>>
>> This will be an elementary question for most but it has me quite
>> roadblocked. Having gone through the entire archive, I could find nothing
>> on
>> this particular subject. Thank all in advance for their patience and
>> support.
>>
>> An entry-level cluster is composed of two Intel WinXP-Pro computers
>> connected via a cross-over cable. The network is configured such that 
>> each
>> computer can see shared files and drives on the other machine. The 
>> printer
>> is shared between the two machines. Each machine can ping the other
>> machine.
>> The MPICH2 installer has been run on both machines. Both machines are in
>> the
>> same workgroup and one logs onto either in the same way.
>>
>> When running the very basic example cpi.exe on either machine, cpi works
>> just fine. MPIEXEC performs as expected. However, when an attempt is made
>> to
>> run across the network, MPIEXEC reports that it can not connect to the
>> other
>> machine. It finds the machine but it can not connect to it. It is here
>> that
>> I am stuck and would appreciate any advice.
>>
>> With Linux, everything works really well. We have had no trouble and a 
>> lot
>> of success.
>>
>>
>> Thanks,
>>
>> Peter.
>>
>>
>
>