[gmx-users] Restart a stopped simulation after a power failure? : Not working in Gromacs 5.1.4

[Mark Abraham]

Hi,

Some of your commands and outputs are inconsistent. You can't get
md_0_1.log without e.g. using -deffnm, and you'll only get md.log if you
don't use deffnm. But if you change that when trying to do your
continuation, mdrun won't know how to append because you haven't told it
that the old name. Be consistent. Your whole trajectory is present in all
your output files, but you're going to have to concatenate them yourself,
e.g. with gmx trjcat since you didn't tell mdrun a consistent name for the
files.

GROMACS 2016 likely refuses to do your continuation, precisely so that you
are prompted to say what you actually want, rather than mdrun trying to be
clever and getting it wrong.

Mark

On Thu, May 18, 2017 at 11:03 AM Adarsh V. K. <adarsh_p130085bt at nitc.ac.in>
wrote:

> Dear gmx users,
>
> I use Gromacs 5.1.4 for protein ligand simulation. How can I restart a
> stopped simulation after a power failure? (used 8 processor cores +
> Graphics card GTX 780Ti).
>
> I have attached the log file along with this mail.
>
> It appeared that no details appended after restarting simulation (after
> power failure at 3.5 ns), Even Though the 'terminal' Ubuntu 16.04 shows
> that the simulation successfully restarted from 3.5 ns (used 8 processor
> cores + Graphics card GTX 780Ti) and completed the entire 8 ns
> simulation...!!. Frames...350 -> 3500 ps (not appended the log file after
> the restart from 3500ps to 8000ps).
>
> I used the command (gromacs 5.1.4),  mdrun -s md.tpr -cpi state.cpt
>
>           # gmx mdrun -s md_0_1.tpr -cpi md_0_1_prev.cpt -v
>
> I also tried, mdrun -s md.tpr -cpi state.cpt -append
>
>
> ------------------------------------------------------------------------------------------------
> md_0_1.log : (final)
> -----------------
>         Step           Time         Lambda
>         1765000     3530.00000        0.00000
>
>    Energies (kJ/mol)
>        G96Angle    Proper Dih.  Improper Dih.          LJ-14     Coulomb-14
>     6.73096e+03    3.13198e+03    2.15931e+03    1.31375e+03    4.51549e+04
>         LJ (SR)  Disper. corr.   Coulomb (SR)   Coul. recip.      Potential
>     3.66659e+05   -7.80170e+03   -2.71529e+06    4.12416e+03   -2.29382e+06
>     Kinetic En.   Total Energy    Temperature Pres. DC (bar) Pressure (bar)
>     4.12474e+05   -1.88135e+06    3.00209e+02   -7.62778e+01    2.10104e+01
>    Constr. rmsd
>     2.37431e-05
>
> -------------------------------------------------------------------------------------------------
> md.log (final)
> ----------
>         Step           Time         Lambda
>         4000000     8000.00000        0.00000
>
> Writing checkpoint, step 4000000 at Thu May 18 13:55:01 2017
>
>
>    Energies (kJ/mol)
>        G96Angle    Proper Dih.  Improper Dih.          LJ-14     Coulomb-14
>     6.44959e+03    3.18212e+03    2.06343e+03    1.53953e+03    4.49878e+04
>         LJ (SR)  Disper. corr.   Coulomb (SR)   Coul. recip.      Potential
>     3.66131e+05   -7.78489e+03   -2.71259e+06    4.78708e+03   -2.29124e+06
>     Kinetic En.   Total Energy    Temperature Pres. DC (bar) Pressure (bar)
>     4.13852e+05   -1.87738e+06    3.01212e+02   -7.59495e+01    5.34060e+01
>    Constr. rmsd
>     2.33151e-05
> ---------
> Force evaluation time GPU/CPU: 12.005 ms/12.263 ms = 0.979
> For optimal performance this ratio should be close to 1!
>
>                Core t (s)   Wall t (s)        (%)
>        Time:   312346.797    39374.380      793.3
>                          10h56:14
>                  (ns/day)    (hour/ns)
> Performance:        9.967        2.408
> Finished mdrun on rank 0 Thu May 18 13:55:02 2017
>
>
> -----------------------------------------------------------------------------------------------
> But while using the command it shows only 3.5 ns of simulation. How to
> solve this problem
>
> gmx trjconv -s md_0_1.tpr -f md_0_1.xtc -o md_0_1_noPBC.xtc -pbc mol -ur
> compact
>
> Select 0 ("System")
>
> -------------------------------------------------------------------------------------------------
>
> Later I repeated the whole 8 ns simulation from beginning.... It completed
> with out any issues.
>
> Can any body tell me what happens during the power failure and why
> re-starting a stopped simulation is not working.
>
> Is it any modification in 're-start' command is required?
>
> Regards,
> Adarsh V. K.
>
>