[herrolsen]

Hey everyone!

I have a problem, fluent is shutting down without a single warning.
I'm running an implicit VOF, implicit body forces, gravity and open channel flow. Everything converges nicely in around 20 iterations.

I'm working on a remote computer, everything looks fine until the window suddenly freezes and I get kicked out of the connection. Me getting kicked out could mean it looses its connection to the net, but no way a 10 sec of lost connection will cause fluent to shut down!

I get no error messages in workbench whatsoever.


Has anyone else had this problem?

[herrolsen]

Managed to do a screenrecord to read the error message.
It says:
Node 2: Process 227548: Recieved signal SIGSEGV.
MPI Application rank 0 exited before MPI_Finalize() with status 2

I read somewhere that people ended up needing to update their version of Fluent to make it work again. I am on 15.0.0 so thats not an option.

Do I have any other options or do I have to restart in serial?

[lisa_china]

you use parallel computation? maybe you can try to change MPI to another transmission protocol.

[herrolsen]

Thanks for answering! Can you point me in the right direction to do so?
Edit: Yes! Running i parallell

[lisa_china]

Originally Posted by herrolsen

Thanks for answering! Can you point me in the right direction to do so?
Edit: Yes! Running i parallell

there is a "parallel setting" button when you start fluent(i use fluent 14.5) and usually i use pcmpi.
and herrolsen, i'm not sure whether it will work.

[HyperNova]

maybe the software need to access to some data in domain that you have not sepecified

[herrolsen]

Originally Posted by lisa_china

there is a "parallel setting" button when you start fluent(i use fluent 14.5) and usually i use pcmpi.
and herrolsen, i'm not sure whether it will work.

Allright, so it means scrapping the data I have and starting over again either way :/
Which MPI type should I pick, intel?

[herrolsen]

Originally Posted by HyperNova

maybe the software need to access to some data in domain that you have not sepecified

Im not sure I understand. Could it be a result of a bad mesh?

[LuckyTran]

I also think it's an mpi problem. To verify, try running in serial mode and see if the problem still occurs.

[HyperNova]

for example in a udf the code needs to access to tempreture but the energy equasion is not turned on

[herrolsen]

definitely! I've tried on two different computers and both shuts down at the exact same time. The fact that they both shuts down at the same time is what led me to believe that it could be a mesh problem. The time is around where the water starts reaching the bottom! Could it be possible that I have a problem in my mesh and this causes the problem?

HyperNova - I do not use any UDF and even left surface tension off, so I think we can rule that out. Thanks for explaining!

[herrolsen]

So far I've tried the different MPIs on available on the computer and only pcmpi (default) works, choosing another one causes fluent to crash. So for now, as suggested, it is running in serial!

[lisa_china]

did fluent also shut down when you using pcmpi?

[herrolsen]

from what I can tell pcmpi is the default one, so this is the one that crashes at the specific time. The other MPIs wont start at all!

[lisa_china]

ok, sorry to mislead you

[herrolsen]

No misleading, my explanation was probably a little fuzzy to begin with I appreciate all help!