[gmx-users] problem with mdrun in parallel

[Mark Abraham]

huifang liu wrote:
> Hello, everybody,
>  
> I recently installed FFTW-3.0.1, mpich-1.2.7 and Gromacs 3.3.3 on my 
> workstation with two 8-cords CPUs.  I think i installed them correctly, 
> beause it runs normally with other parellel MD software. It also went 
> nomorally when i  run mdrun command in gromacs with 4 nodes. The problem 
> is it doesn't run with when it is up to 8 nodes. In fact, i only change 
> the '-np 4' to '-np 8' in both grompp and mdrun commmand. When i typed 
> the 'top' command, it shows there are only two node are running. And 
> after a little while, it stopped run with the follow error:
> p7_19858:  p4_error: Timeout in establishing connection to remote process: 0
> p5_19808:  p4_error: Timeout in establishing connection to remote process: 0
> p6_19832:  p4_error: Timeout in establishing connection to remote process: 0
> p0_19710: (324.359375) net_recv failed for fd = 9
> p0_19710:  p4_error: net_recv read, errno = : 104
>  
> In addition, it runs very well, when i use '-np 5'

So how is your MPI environment configured?

Mark