GROMACS can't find files that are in the working directory

User discussions
1

GROMACS version: 2022.4
GROMACS modification: No

Hi all,

I am running umbrella sampling on an aggregated peptide filament. I have a fairly large box and am timing out on the computing cluster I’m using. Originally, I thought no worries, I can just continue the job. I made a mistake and reran the job instead of continuing it. This generated the usual batch of #file# backup files. I renamed the new files with the older #file# files because they had more progress into the simulation. I did this intentionally to overwrite the new files.

I ran gmx mdrun -deffnm umbrella362 -cpi umbrella362.cpt -px umbrella362_pullx.xvg -pf umbrella362_pullf.xvg

And got the following error:

Checksum wrong for ‘umbrella362.log’. The file has been replaced or its
contents have been modified. Cannot do appending because of this condition.

I have double-checked that the file is present (a different error arises when I move umbrella362.log out of the working directory) and I haven’t modified the contents of the file, only the name of it. Given I’ve confirmed both of these things, I don’t know why GROMACS can’t find/read the file. Is there a way to undo whatever change I made to the umbrella362.log file?

2

Your issue is likely that you renamed old files but still have the new checkpoint file, so there is a mismatch. I suppose your rerun is so long that it also overwrote the previous checkpoint file.

If you have a trr file you could run grompp to generate a new tpr to continue from.

3

You’re correct. I still have the old .log files, but the .cpt files got overwritten by the new job. I sadly don’t generally write .trr files for the sake of storage space and will have to redo some simulations, but I’m glad to know what’s going on.

Do you know if there’s a reason the old .cpt file doesn’t get saved in the #file# format like the rest of the files, or if there’s an option to prevent overwriting of .cpt files?

In any case, thanks for your response, Dr. Hess.

4

Good question. We never thought about this. When appending there is no point. But when not appending we should also back up the checkpoint files.

I filed an issue for this: Also back up checkpoint files (#5494) · Issues · GROMACS / GROMACS · GitLab