[gmx-users] Continue running in Gromacs 4.0

[Mark Abraham]

xuji wrote:
> Hi all:
> 
> First I should say sorry for all the people who have read this thread for the same following content as the thread
> with wrong title I wrote yestoday "cCntinue running in Gromacs 4.0" .

That's OK - people will read around a typo, particularly when you've
tried hard to give a full description of your problem :-)

> In Gromacs-4.0 manual it's said that we only need the check point file to continue the simulation.
> I have a simulation with 16 threads at two nodes. I used the following command to run the simulation:
> 
> mpiexec -machinefile ./mf -np 12 mdrun -v -cpo dppc_md_1.cpt -cpt 60 -s dppc_md.tpr -o dppc_md.trr -x dppc_md.xtc -c dppc_md.gro -g dppc_md.log -e dppc_md.edr >& dppc_md.job &

It's abnormal to lack an _mpi suffix, but I presume you know for sure
that your mdrun was configured properly with MPI...

> but when I continue the simulation with 
> 
> mpiexec -machinefile ./mf -np 12 mdrun -v -cpi dppc_md_1.cpt -cpo dppc_md_2.cpt -cpt 60 -append -s dppc_md.tpr -o dppc_md.trr -x dppc_md.xtc -c dppc_md.gro -g dppc_md.log -e dppc_md.edr >& dppc_md.job &
> 
> The mdrun program stopped with "No previous checkpoint file present, assuming this is a new run."

OK that probably means what it says. Your -cpt option told mdrun to
write a checkpoint every 60 minutes. If your first run didn't go that
long, then that might be the problem. I'd also expect GROMACS to write
the checkpoint file on a normal exit, however. You should check what
files are present and how large they are and what their permissions are.

Also, you should try simplifying things until you work out what the
problem is... so experiment with a run with few integration steps,
observing the effect of -append, the -cpi/-cpo/-cpt options, on
different runs to see where the problem might lie.

> I checked the files in the simulation directory, there is also a file called "dppc_md_1_prev.cpt".
> So I try the file called "dppc_md_1_prev.cpt" with 
> 
> mpiexec -machinefile ./mf -np 12 mdrun -v -cpi dppc_md_1_prev.cpt -cpo dppc_md_2.cpt -cpt 60 -append -s dppc_md.tpr -o dppc_md.trr -x dppc_md.xtc -c dppc_md.gro -g dppc_md.log -e dppc_md.edr >& dppc_md.job &
> 
> I also got "No previous checkpoint file present, assuming this is a new run."	
> So can someone tell me what's wrong with my method and how can I continue a simulation in Gromacs-4.0?