MPI errors on a linux cluster (ABAQUS/Explicit)

Hello everybody,

i try to perform a 2D orthogonal cutting simulation using the Johnson-Cook damage model in ABAQUS/Explicit. When simulated on my desktop PC (AMD Phenom quad core processor, Windows Vista 64bit), the simulation runs successfully. But if  I try to run the simulation on a linux cluster, it ends with error messages related to HP-MPI. Moreover, if I run a 3D drilling simulation on the cluster, it runs successfully with no errors.

I use ABAQUS 6.7.5, CPE4R elements. 

Example of error messages from *.log file:

ABAQUS/Explicit checked out 15 tokens.

<0 out of 15 licenses remain available>.

forrtl: error (75): floating point exception

MPI Application rank 10 killed before MPI_Finalize() with signal 6

forrtl: error (78): process killed (SIGTERM)

forrtl: error (78): process killed (SIGTERM)

forrtl: error (78): process killed (SIGTERM)

forrtl: severe (174): SIGSEGV, segmentation fault occurred

Image PC Routine Line Source

libirc.so 00002B1624CA2CEA Unknown Unknown Unknown

libirc.so 00002B1624C9D606 Unknown Unknown Unknown

libifcoremt.so.5 00002B16256BBD7B Unknown Unknown Unknown

forrtl: severe (174): SIGSEGV, segmentation fault occurred

Image PC Routine Line Source

Floating point exception does not occur in every simulation (I think it occurs only when a mesh i denser, i.e. 0.02 mm), MPI Application rank errors occurs in every simulation (except 3D simulation as mentioned above). 

Cluster configuration is as follows:

OS: CentOS release 5.5 (Final) 64bit Kernel 2.6.18-194.3.1.el5 x86_64

HW: 2 x IBM blade HS21 (Type 8853) in BladeCenter-H

HW blade: 2 x Intel(R) Xeon(R) CPU E5440 @ 2.83GHz 8 GB RAM

Interconnect: 1Gbit ethernet

 

Could you post any advice?

Regards, M. Madaj.