Fatal error: There is no domain decomposition

User discussions
1

GROMACS version: 2019.1
GROMACS modification: No
Here post your question: I’m writing, because I’m a kind of new gromacs user, thus, I’m trying to run a calculation (in between many more) in a cluster using the following slurm .sh file:

#!/bin/bash
#SBATCH --nodes=20
#SBATCH --ntasks-per-node=10
#SBATCH --cpus-per-task=4
#SBATCH --mem=0
#SBATCH --time=0-00:20
module load intel/2018.2 intelmpi/2018.2 gromacs/2019.1
export OMP_NUM_THREADS="${SLURM_CPUS_PER_TASK:-1}"

srun gmx_mpi mdrun -deffnm md -s topol.tpr

However all the time I get the following error message:

Fatal error:
There is no domain decomposition for 112 ranks that is compatible with the
given box and a minimum cell size of 1.27375 nm
Change the number of ranks or mdrun option -rdd or -dds
Look in the log file for details on the domain decomposition

I generate the tpr file with the command:

gmx_mpi grompp -f md23.mdp -p agua-sds23.top -c inicial23.pdb

I read the manual, but I still can not find what am I missing. Would you mind helping me? What are we missing?

All the best and thank you in advance,

Eduardo Schott

2

You are attempting to run the simulation across too many processors.

3

I understand that, but how can I get full usage of the resources? Is there any option?

Thanks!

4

The solution is to use fewer nodes/CPUs. Your system is too small to be split across so many processors.

5

If the goal is to use all the resources, make a bigger simulation box. (and this is unlikely to be what you actually want to do)

If, however, the goal is to run your given simulation box at as fast as possible then you need to some short simulations with different numbers of processors to find the fastest configuration. You should do this at any time you are starting on a new HPC, simulation size, software version etc. Then you ensure you use your provided resources efficiently, rather than wastefully.