[febriyan91]

Dear Foamer.

I am running OpenFOAM-v2012 on dual-CPU Computer (2x Xeon E5 @14 cores=28 thread, so total 56 threads on the Computer). I am researching with rhoCentralFoam to run 2D axisymmetric case. The domain decomposition works well after I run "decomposePar". But the problem arose when i tried to run the case in parallel with all threads.

I used this command to run it on script named ./gas0 , "mpirun -H localhost --oversubscribe -np 56 rhoCentralFoam -parallel > log.run1"
then I run "tail -f log.run1". but the terminal looks freeze as shown in picture 1. Then i run "top" on other terminal and there were 56 process running rhoCentralFoam. I check the System Monitor also the CPUs were full occupied.

Here are the additional information:

1. I have tried install openfoam from precompiled and built from source but the same problems arose.
2. The computer is dual boot between windows10 and ubuntu 20.04.
3. gcc version 9.3.0, g++ version 9.3.0, mpiexec --version 4.0.3
4. the simulation runs propely on SERIAL.
5. the simulation works properly when the processor assigned not exceeding 14. "-np 14".
6. I could upload the case if necessary.

I am really newbie running and installing OpenFOAM on dual-nodes processors/server PC/small cluster. Previously I run OpenFoam on single node CPU, this type of problem never arise. I have searched some references in this forum but couldnot find the solution. I will be very thankful if someone could guide me to find the solution to overcome this problem to runs openfoam parallel all cores on dual-nodes PC. Any guidance to reference or forums threads are very welcome x)

Regards, Febriyan

[gmig]

Hello Foamers,

Bringing this up with a similar question of my own

I have been running OpenFOAM (v2112 now) for some time on a 8 core processor without issues when working in parallel.

I have now access to a 2x CPU machine with 2 x Intel® Xeon® Gold 6240 18c. I wanted to give it a try to further decompose my domain and check how it compares with my previous setup.

I have the same problem as the OP. I have been able to decompose up to 18c but I cannot get it to go any further without triggering the error that is discussed above.

I have similar specs (mpiexec --version 4.0.3...)

I think mpi only "sees" one processor and that I should probably specify that I have in fact 2 CPUs.

After some investigation I have tried running:

mpirun --host node1:16,node2:16 -np 32 snappyhexmesh -parallel

with no success, I get the following:

ssh: Could not resolve hostname node2: Temporary failure in name resolution
 -------------------------------------------------------------------------- 
ORTE was unable to reliably start one or more daemons. 
This usually is caused by:...

and

ORTE does not know how to route a message to the specified daemon
located on the indicated node:

Any help will be appreciated!

[Jasper Z]

Originally Posted by gmig

Hello Foamers,

Bringing this up with a similar question of my own

I have been running OpenFOAM (v2112 now) for some time on a 8 core processor without issues when working in parallel.

I have now access to a 2x CPU machine with 2 x Intel® Xeon® Gold 6240 18c. I wanted to give it a try to further decompose my domain and check how it compares with my previous setup.

I have the same problem as the OP. I have been able to decompose up to 18c but I cannot get it to go any further without triggering the error that is discussed above.

I have similar specs (mpiexec --version 4.0.3...)

I think mpi only "sees" one processor and that I should probably specify that I have in fact 2 CPUs.

After some investigation I have tried running:

mpirun --host node1:16,node2:16 -np 32 snappyhexmesh -parallel

with no success, I get the following:

ssh: Could not resolve hostname node2: Temporary failure in name resolution
 -------------------------------------------------------------------------- 
ORTE was unable to reliably start one or more daemons. 
This usually is caused by:...

and

ORTE does not know how to route a message to the specified daemon
located on the indicated node:

Any help will be appreciated!

Dear Ignacio,

Have you found the solution? I have the same problem. I have a dual-cpu with 24*2 cores, but it seems that I can only use one cpu when I use the general script mpirun -np 38 interFoam -parallel. It will give me errors when I use cores larger than 24. How to make parallel runs on 2 cpus? Any suggestions?

Thank you,
Jasper

[fardin.sh]

Dear all,


febriyan91, he made the correct code and for sure his case at that time was running.
What he had do was just delete "> log.run1" because this command says to just write output to a file named "log.run1". So, literally there wont be anything printed in the terminal.



But If you want to use all cores of a multi-CPU PC you need to run your solver using following command:


"mpirun -H localhost --oversubscribe -np 56 rhoCentralFoam -parallel"


where 56 is your number of cores.