[fxzf]

Hello Foamer,

I got a strange issue.

My case is running fine in sequence mode, but if I decompose it and run in parallel. it just crash after a few iterations with U goes crazy.

I tried a few times to use different decomposition method, and in oscationally with certain decomposition settings, it can be run in parallel.

What is the best way to solve this?

Thanks

[ajithnair]

does mpi run show any error message. Or is it crashing like a CPU deadlock where one CPU is waiting for the other.

check the processes by using 'top' command.
'mpirunDebug' may also help or you can write the run to a logfile and use 'tail -f 'filename'' to see what the error is

[fxzf]

Originally Posted by ajithnair

does mpi run show any error message. Or is it crashing like a CPU deadlock where one CPU is waiting for the other.

check the processes by using 'top' command.
'mpirunDebug' may also help or you can write the run to a logfile and use 'tail -f 'filename'' to see what the error is

No, it crash like U went to crazy value, so temperature will get negative. No any error throw from mpi side.

but if I just reconstruct the mesh and run in serial, it works and no velocity diverge happens.

[fxzf]

Originally Posted by ajithnair

does mpi run show any error message. Or is it crashing like a CPU deadlock where one CPU is waiting for the other.

check the processes by using 'top' command.
'mpirunDebug' may also help or you can write the run to a logfile and use 'tail -f 'filename'' to see what the error is

Ok, seems found problem. Some value left in processor boundary which screw the temperature and velocity equation. That's why running in serial mode no this issue.