[gmx-users] Problem of Gromacs Run on MPICH cluster.

[PeiQuan Chen]

Dear gmx-users,

I have compiled successfully the parallel version of gromacs in a Linux Cluster implemented with MPICH and PBS batch queue system.

I can successly run the parallel mdrun when I use the interactive mode of PBS with the
following command:
qsub -l nodes=<number of nodes> -I
and then run the parallel mdrun with the mpirun command:
mpirun -np $NP -machinefile ${PBS_NODEFILE} mdrun -np $NP -s myfile.tpr
This model run successfully.

But when I come to use the PBS script files to submit the job，the PBS system complaint that:
    p4_error: latest msg from perror: Resource temporarily unavailable
p0_11046:  p4_error: net_accept accept: -1
Killed by signal 2.^M
p0_11046: (4.112234) net_send: could not write to fd=3, errno = 32

PBS scritp files:
#PBS -l nodes=2:ppn=2:enet
#PBS -N my_job_name
#PBS -j oe
NP=`wc -l ${PBS_NODEFILE} | awk '{ print $1; }'`
cd My_working_directory
mpirun -np $NP -machinefile ${PBS_NODEFILE} mdrun -np $NP -s myfile.tpr

Could anybody tell me how to resolve this problem? 

Thanks in advance.


Sincerely yours
---- 
PeiQuan Chen
________________________________________________________________________
PeiQuan Chen 	Department of Chemistry, Nankai University
WeiJin Road 94, TianJin, 300071, P.R.China
phone:	(86)2223507510		Mobile Phone: (+86)13920675030
gromacs at 163.com	
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++