[gmx-users] Gromacs 4 bug?

[Bernhard Knapp]

Dear users

I have a problem with gromacs 4 (also tried 4.0.2). If I start a 
simulation it is calculating for some time but after about one day it 
gets somehow stuck. This means that the simulation is still running and 
the top commands says md still uses all 4 cores of the computer:

5487 bknapp    20   0  123m  41m 6864 R  100  2.0   1388:25 
mdrun                                                                                                              

 5488 bknapp    20   0  114m  35m 6432 R  100  1.8   1388:33 
mdrun                                                                                                              

 5489 bknapp    20   0  114m  35m 6432 R  100  1.8   1388:00 
mdrun                                                                                                              

 5490 bknapp    20   0  113m  34m 6424 R  100  1.7   1388:16 mdrun

but no output is written anymore. I redirect the output of std::err to 
job.1sjh.md.err (which contains then the steps done and the prediction 
how long it will take to finish):

-rw-r--r-- 1 bknapp users     258713 2008-12-12 06:25 job.1sjh.md.err
bknapp at quoVadis01:~/gromacsRuns/_validation> date
Fri Dec 12 09:30:12 CET 2008

It seems that it is still calculation (something?) but not writing 
output anymore. In this case it stops after 423900 steps or 847ps. If I 
simply re-run the same simulation it stops again but at a different 
time. If I perform exactly the same commands on gromacs 3.3 it works 
fine (I stopped the simulation by myself after 20ns).

Is this a known bug or is there a work-around?


cheers
Bernhard