[gmx-users] fatal errors in parallel but not on single processor

[David van der Spoel]

Patricia Francis-Lyon wrote:
>   Hi, I'm new to gromacs and I can't figure out what I'm doing wrong 
> with parallel runs. I'm running gromacs 3.3.2 on a 215 residue protein: 
> 1rz4.pdb. The 2 missing residues were added with modloop followed by 
> energy minimization. The protein was put in a dodecahedron box and 
> solvated with a layer of water at least 1.2 nm from the protein and 14 
> Na+ ions were added to bring the total charge of the system to zero. 
> Energy minimization was done to remove close contacts.
>  
> When I run on a single processor my MD runs seem fine (heatup position 
> restrained MD, constant temperature position restrained MD, and now the 
> production MD run is almost 86% complete).  But when I run on a cluster 
> some I get strange fatal errors, maybe tripped by boolean checks during 
> writing of energy info to file. During position restrained MD I got:
> fatal error in function do_enx (line 343 of enxio.c)
> during the call:
> gmx_fatal(FARGS,"could not write energies")
> The fatal error happened after printing Coulomb1-4 energy in the log 
> file, the next line of headings was NOT printed.
>  
> During production run of MD I got:
> Source code file: stat.c, line: 283
> Fatal error: XTC error

This error usually means that either your box explodes, or your disk is 
full. Could it be either of these problems?

If not, is it reproducible?


> happening after writing these lines in the log:
>            Step           Time         Lambda
>          257600      515.20001        0.00000
>  
> These errors do not appear when I run equilibriation and production MD 


-- 
David van der Spoel, Ph.D.
Molec. Biophys. group, Dept. of Cell & Molec. Biol., Uppsala University.
Box 596, 75124 Uppsala, Sweden. Phone:	+46184714205. Fax: +4618511755.
spoel at xray.bmc.uu.se	spoel at gromacs.org   http://folding.bmc.uu.se