mpirun hangup.. can anyone help how to fix mpirun issues.
 

Hi,

I am getting hangups in my mpirun randomly.

The error message i am getting is


"mpirun noticed that process rank 18 with PID 28060 on node amd32-1 exited on signal 1 (Hangup).
--------------------------------------------------------------------------
2 total processes killed (some possibly by mpirun during cleanup)".

Does anybody see this kind of error before?

Any ideas of what needs to be done to resolve this issue.

I appreciate any suggestion.

Thanks in advance.

If you try to run the solver after logout, you should start your job like:
(nohup=no hangup)

hi,
I dont know what is your problem exactly! you didnt specify in which situations and when in happens; but parallel running of cases is somewaht crashy at all. I havent experience OpenFOAM crash with mpirun but a lot with FLUENT parallel running. in FLUENT cases, the problem was with my system's RAM. it is better to place same type and capacity RAM in all slots of MB, but if you dont have, be aware of slot numbers and obey it to place them. after that a memtest may be useful to ensure that the problem is or is not with your RAM.

may can help,

Hi pit,

Thank you for the reply.

I am using nohup command to run the simulation in background.

It seems like it is something related to memory , if I ma not wrong.

Hi Montazari,


I am running a boundary layer simulation with cyclic boundary conditions.
I am running it parallel on 26 processors.
This error message occuring randomly.
I ahve enough RAM capacity.
I am new to OpenFOAM, I didn't get that "memtest", Could you please explain me?

Thanks&Regards.

Hi,

Which version of OF you're running?

I have run cases with periodic boundary conditions and metis decomposition using the following on the decomposeParDict:

preservePatches (recycle_1 recycle_2);

where recycle_1 and recycle_2 are my cyclic patches.

good luck,

Tony

Hi Tony,

I am using OF version 1.7.

I have many different cyclic patches as my mesh is not uniform.
I have 16 spanwise and 46 streamwise cyclic patches. How do I can define them in metis decomposition?

I used simple decomposition for my domain.

Hi,

I am using 1.7.1 as well,

I have only done this with simple domains (box) and I use the following in my decomposeParDict:

method metis;
preservePatches (recycle_1 recycle_2);

my patches are named recycle_1 and recycle_2

Your patches would go inside preservePatches()

gl,

Tony

Thanks Tony,

I will try with metis decomposition.

Thanks alot.

Regards,
Pam.

hi pam,
I recommend checking RAM slots and their preferred arrangement of slots on motherboard first of all; then use memtest86+ utility to check memories thoroughly. memtest86+ utility is available in some distributions of linux like ubuntu by default and is available to download for other distributions in its official site. (the utility can be selected in grub menu in ubuntu if it hasnt turned off manually)

Hi all,

As I found one suggestion as changing the "minimum buffer size" in openfoam/etc/settings.sh file. I modified that and running now...
Till now it is running well.. But not sure how that will be helpful :).

Hi pam
i have the same problem.did this solution work for u? and what should I do with "minimum buffer size"? should I rise it?

Hello everyone,

Have you got any solution for this error?

Regards,
Hossein