Errors in Parallel Processing

[zanakarimi]

I installed latest parallel version of OpenSees. I also tried smpd -status and it returns: smpd running on My-PC.
Now, when I try the following command to run my script:
mpiexec -np 2 OpenSeesSingleParallelInterpreter main.tcl
I get the following:

launch failed: Create Process<OpenSeesSingleParallelInterpreter main.tcl> on 'My-PC' failed, error 2 - The system cannot find the file specified.
child killed: unknown signal

I have OpenSeesSP.exe and main.tcl in a folder on one of my drive. I execute OpenSeesSP.exe from the same folder but don't know what is the reason of the error.

I use tcl/tk 8.5.12

Anybody had such experience with parallel version? I really need help
Thanks in advance

[fmk]

ther .exe is OpenSeesMP not OpenSeesParallelInterpreter

[zanakarimi]

Thanks for your prompt response
I also tried these two commands in OpenSeesMP.exe:
mpiexec -np 1 OpenSeesMP ex1.tcl
and
mpiexec -np 1 OpenSeesManyParallelInterpreters ex1.tcl
I get this for both of them:

Connect on sock <host=????, port=8676> failed, exhausted all end points
Unable to connect to '????',
sock error: Error = -1
child killed: unknown signal

my .tcl file is:
set pid [getPID]
set np [getNP]
puts "Hellow World Process: $pid"

One thing more, when I try ''getNP'' command it returns 1 but I'm using core i7 processor!
smpd -query --> returns:
querying hosts in the default domain: Not implemented.

I appreciate your help in advance,

[fmk]

the NP = 1 is to be expected if you run with np = 1 (i.e. one processor)

a google search indicates the problem maybe how you installed mpich2

look at these threads to see how to install it as administrator.

[reagan]

Hi Frank

Running into the same problem with a set of new MPICH2 installations.

Followed the same procedure as mentioned in http://opensees.berkeley.edu/wiki/index ... Processing. Old installations using the same procedure still work fine. New installations produce a "launch failed" error when trying to run OpenSeesMP. Trying to execute the cpi.exe executable in the MPICH2 example folder produces no errors though. Only difference is that probably the old installations are running an older version of tcl: 8.5.12, while new installations are running tcl v8.5.14.

On a different note, all links to OpenSeesParallel, tcl/tk, and MPICH2 from http://opensees.berkeley.edu/OpenSees/p ... wnload.php are dead.

Reagan

[reagan]

Just fixed it. It was a problem with the user credentials stored in the registry. Removed and re-stored the credentials to get it working.

[igidaris]

Hi Reagan

I have the same problem as you had : cpi.exe works but when I try to use mpiexec with opensees (e.g: mpiexec -np 4 OpenSeesMP test.tcl) I get the error msg :"Child kiled:unknown signal".

I tried to remove my credential using mpiexec -remove and then restore them using mpiexec -register but I still have the same problem!

Could you please explain in more detail if you have done anything else?

I would greatly appreciate if anyone knows what exactly the problem is and how I can fix it.

[reagan]

This was a long time ago. I don't remember the details, but I removed my user credentials from the registry by running wmpiconfig.exe and then re-registered them the same way I did when I first installed MPICH2.