[gmx-users] question_gromacs

[Justin Lemkul]

On Sun, Dec 15, 2013 at 3:20 AM, Andrew Bostick
<andrew.bostick1 at gmail.com>wrote:

> Dear Gromacs users
>
> I have following error when I do equilibration:
>
> There is no domain decomposition for 4 nodes that is compatible with
> the given box and a minimum cell size of 0.8875 nm Change the number of
> nodes or mdrun option -rcon or -dds or your LINCS settings.
>
> I tried all suggestios in above error and Error page ogf Gromacs web site:
>
> 1) I changed -rcon and -dds and LINCS settings, but I have same fatal
> error, again.
>
> What is the optimum or the best values for -rcon and -dds and LINCS
> settings in this case (error)?
>
>
The default DD settings work in nearly all cases.  Some systems are simply
too small to be parallelized effectively and DD simply won't work.  I
suspect this is the issue in your case, though you haven't told us much
about what the system is or how big it is.


> 2) I used -nt 1, this time, I had not error, but I encountered:
> step 0Segmentation fault.
>

An immediate segmentation fault indicates that the initial geometry,
topology, or run parameters are bad.  In the case of your run parameters,
some things stand out:


>
> -----------------------------------------------------------------------------------------------
> My mdp file is as follows:
>
> title               =  CNT_H
> constraints         =  all-bonds
> constraint_algorithm = LINCS
> integrator          =  md
> dt                  =   0.001    ; ps
> nsteps              =  30000  ; total : 30 ps .
> nstcomm             =  1
> nstxout             =  500
> nstxtcout           =  500
> nstvout             =  500
> nstfout             =  0
> nstlog              =  500
> nstenergy           =  500
> nstlist             =  10
> ns_type             =  grid
> rlist               =  0.4
> rcoulomb            =  0.4
> rvdw                =  0.4
>

These cutoff values make no sense and are certainly a source of
instability.  The cutoffs are an important component of the force field and
cannot be changed according to system size.  The fact that you have had to
reduce these to such nonsensically small values tells me your system is
very small and grompp was probably warning about minimum image violations,
hence also why DD cannot decompose the system adequately.


> Tcoupl              =  V-rescale
> tc-grps                =  (NU   HHM
> tau_t               =  0.2   0.2
> ref_t               =  77    77
> energygrps            = (NU    HHM
> Pcoupl              =  no
> compressibility     =  4.5e-5
> gen_vel             =  yes
> gen_temp            =  100.0
>

You're generating velocities at 100 K, then trying to use 77 K as a
reference temperature.  Though V-rescale is rather robust, you're not
equilibrating in the most stable of ways by doing this.


> gen_seed            =  173529
> pbc                       = xyz
> coulombtype         = PME
> vdwtype              = Cut-off
> fourierspacing = 0.12
> ewald_rtol = 1e-5
> freezegrps           = (NU     HHM
> freezedim            = Y Y Y   N N N
>

If you are not freezing group HHM, it is unnecessary here (though
immaterial to the problem).  Freezing is a serious perturbation and can
actually inhibit equilibration in some cases; is it really necessary to do
this?

-Justin

-- 

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

Justin A. Lemkul, Ph.D.
Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 601
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441


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