mdp option simulation_part does not work as expected
1. If I add the line "simulation_part=4" to a .mdp file and then start that run when there is no checkpoint file, the output files do not contain a .part####. extension.
2. Running a restart of this run (using the .cpt file) yields the .part0005. segments.
I realize that this does match up with the exact wording of the description from mdrun -h
-[no]addpart bool yes Add the simulation part number to all output
files when continuing from checkpoint
but then I'm not sure what the point of allowing simulation_part=4 is in the .mdp file.
In my case, I experienced a few hours worth of .xtc data corruption. I am rerunning the trajectory from the last good .xtc frame that I have via grompp as I no longer have the .cpt files from that point. It would be nice if I could simply set the simulation part and everything would carry on. This, however, is not possible.
I guess that I could generate the .cpt by doing a zero-step mdrun first with simulation_part=(N-1), but that's a bit cumbersome when running many simulations via scripts.
#1 Updated by Berk Hess over 9 years ago
I think this issue has been nearly resolved already.
In git mdrun -append is default (with -cpi).
-append now does one of the following three things:
1) if no output files are present, write new files (without part????).
2) if all output files are present and all checksums match those
now stored in the checkpoint file, append.
3) otherwise generate a fatal error.
Thus you can no longer accidentally corrupt or truncate old files.
I would suggest to remove the mdrun -addpart option and only
but always add .part???? with -noappend.