[gmx-users] Domain decomposition error while running coarse grained simulations on cluster

[Avijeet Kulshrestha]

>
> Here's your problem. You have pairs defined that are in excess of 12 nm,
> but they are assigned to a 1-4 interaction, so atoms that should be
> separated by three bonds. The user-defined potential shouldn't matter
> here unless you've added [pairs] to the topology.
>
> I see your point.
What can I do to define LJ potential between two atoms which are actually
far apart?
and the results from the single processors are correct?
I tried it in a different way too,
I generated tables between pairs which are far apart (with the LJ
potential) and kept them in the bonded section with function type 8 (to
exclude LJ from usual force field).
The system has 12282 number of particles and the size of the box is
9.76580  11.92760  11.62620.
I minimized energy with 10 processors and -rdd 1.4. It worked.
In NVT equilibration I am getting the following error:

Atom distribution over 8 domains: av 1535 stddev 38 min 1508 max 1582
Constraining the starting coordinates (step 0)
Constraining the coordinates at t0-dt (step 0)
Not all bonded interactions have been properly assigned to the domain
decomposition cells
A list of missing interactions:
          Tab. Bonds of    818 missing     16
-------------------------------------------------------
Program:     gmx mdrun, version 2018.6
Source file: src/gromacs/domdec/domdec_topology.cpp (line 240)
MPI rank:    0 (out of 8)
Software inconsistency error:
Some interactions seem to be assigned multiple times