[gmx-users] Re: any chance to write avoid writing out trr but only xtc file ?

Justin A. Lemkul jalemkul at vt.edu
Wed Dec 2 14:51:10 CET 2009

Jörn-Benjamin Lenz wrote:
> Hi,
> here is what the is written out if I remove on eof the trr files during the REMD simulation although the xtc files should still be written out.
> just for your information: I am not out of quota. that I can promise !!!
> Any ideas why the program behaves like this ? 

Yes, because in trnio.c, an error is raised if the trajectory file is missing. 
The error message was introduced in version 4.0.3.

> Is there anything that was changed in gromacs which could explain this ?
> Another additional info on the problem: 
> I used the older gromacs versions with lam mpi. now i compiled it against mpich2. Could that be the problem ? 

This has nothing to do with the MPI implementation.

> If so, does anyone have an idea how to circumvent it, i.e. still being able to write out xtc files while the trr files are deletetd and the job does not crash ?

You can't proceed by simply deleting the .trr.  Use Xavier's suggestion about 
.mdp settings.  Setting nstxout, etc = 0 will suppress all .trr output until the 
final frame, which is always written.


> Cheers and many thanks,
> Joern
> [...]
> step 5600, will finish Thu Dec 10 09:43:05 2009
> step 5700, will finish Thu Dec 10 09:41:10 2009
> step 5800, will finish Thu Dec 10 09:39:19 2009
> step 5900, will finish Thu Dec 10 07:51:36 2009
> step 6000, will finish Thu Dec 10 07:51:39 2009
> -------------------------------------------------------
> Program mdrun_mpi, VERSION 4.0.5
> Source code file: trnio.c, line: 252
> File input/output error:
> Cannot write trajectory frame; maybe you are out of quota?
> -------------------------------------------------------
> "Make the Floor Burn" (2 Unlimited)
> Error on node 0, will try to stop all the nodes
> Halting parallel program mdrun_mpi on CPU 0 out of 5
> gcq#181: "Make the Floor Burn" (2 Unlimited)
> [cli_0]: aborting job:
> application called MPI_Abort(MPI_COMM_WORLD, -1) - process 0
> mdrun07 REMD finished
> ln: accessing `traj0.trr': No such file or directory
> ln: accessing `traj1.trr': No such file or directory
> ln: accessing `traj2.trr': No such file or directory
> ln: accessing `traj3.trr': No such file or directory
> ln: accessing `traj4.trr': No such file or directory
> --------------mpdallexit------------------
> Jörn-Benjamin Lenz wrote:
>> thanks you xavier and justin for your answers on my problem, however, i still
>> wonder why i could completely delete the e.g. five trr files (from an REMD
>> simulation on our cluster) named trajX.trr (with X = 0..4) and the
>> corresponding xtc files (trajX.xtc) were still written out in one of the
>> former gromacs versions. now this is not possible any more, i.e. as soon as i
>> delete one of the trajX.trr files, the whole simulation breaks. does anyone
> Do you get any sort of error?  I would think mdrun would at least report the I/O 
> problem.  I don't know what the difference would be in this regard between the 
> older and newer versions, but perhaps some changes were made such that mdrun 
> expects all files standard output files to be present (perhaps to allow for the 
> use of checkpointing?).  Just a guess, but certainly Xavier's solution is much 
> more elegant.  My approach is usually to save very few frames (nstxout, etc = 
> large # of steps), but certainly setting these parameters to zero is even better!
> -Justin


Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080


More information about the gromacs.org_gmx-users mailing list