[gmx-users] Equilibration fails only when run from script

[Justin Lemkul]

On 12/1/15 6:39 AM, R Corey wrote:
> Hello
>
> I am running a series of FEP MD simulations using Gromacs 5.0.4, for each a
> steepest descents minimization, NVT equil, NPT equil and production run,
> all controlled by a central script. However, only about 50% of the runs get
> through the NVT equilibration without crashing - crash error as below:
>
> Fatal error:
> 3 particles communicated to PME rank 0 are more than 2/3 times the cut-off
> out of the domain decomposition cell of their charge group in dimension y.
> This usually means that your system is not well equilibrated.
>
> However, the weird thing is that if I re-run the script the next day,
> choosing only failed jobs to rerun, some of them now work fine. And if I
> simply mdrun each tpr from the command line it runs fine too.
>
> Is this expected, that a tpr file for an NVT equilibration will crash one
> day but run fine the next?
>

This suggests transient instability that arises due to differences in order of 
operations, inherent chaos in MD, etc. If it's that difficult to get running, 
you probably have a larger issue that only shows up transiently. In any case, 
you can always try running with the latest version (5.1.1) in case of a bug.

-Justin

-- 
==================================================

Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 629
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441
http://mackerell.umaryland.edu/~jalemkul

==================================================