Incomplete Simulation Time

GROMACS version: 2018.3
GROMACS modification: No


I tried running npt production for 10 ns, so used the following time step settings:
nsteps = 5000000 ; 2 * 5000000 = 10000 ps (10 ns)
dt = 0.002 ; 2 fs

But the simulation was complete up to only 7 ns. What do you think might have caused this issue?

Check the log files for information, both the .log which is output from mdrun and the output file if you are using something like SLURM.


Thank you for the reply. It says the job ended “due to time limit”.
So I was thinking of just getting rid of the “–time” command from my .sbatch file. If I do that, the simulation will run until it’s done, right?

In your submission details, you should set the time limit to the longest allowed on your cpu/gpu.
And if the run stops again due to time limit, you just need to restart the simulation from the last step performed (mdrun -rerun).

Before starting a production run you should do some short ones to optimise the settings used and determine how long your simulation should take. Then within your SLURM file you specify that time length with a buffer.

By removing the time variable specified within your SLURM file, you will default to the maximum allowed time on the HPC for that type of job. Whether you job is completed in that time, no idea, you will need to do some testing to see how long it should take.

Thank you for the reply. I didn’t know I could resume the simulation from where it stopped.

Thank you for the advice. I should do that in the future.

Finally, the simulation just finished running, and I have been trying to run DSSP analysis.
So I typed: gmx do_dssp -f file.xtc -s file.tpr -o dssp.xpm -sc scount.xvg.
But I got an error saying, “DSSP executable (/usr/global/whatif/dssp/dssp) does not exist (use setenv DSSP)”.

How do I resolve this problem? When I was in college a few years ago, I didn’t get this error (probably because it was the older version of Gromacs and DSSP?).

Start a new thread for this new, un-associated question.


So when I submit the job to resume the simulation, is this the only command that’s needed (mdrun -rerun), or do I write mdrun -rerun npt_pro for example?

You do not want to use mdrun -rerun, which is for re-evaluating energies from an existing trajectory. What you want is to restart/resume your simulation from the checkpoint file. See