[gmx-users] simulation crash

[Andrea Spitaleri]

Hi there,

I am running several independent replica of similar system but in 
different pose of protein-peptide. I am in NVT ensemble and the length 
of the simulation is 125ns (dt=2fs). Few of them are crashing with the 
classic error:

"1 particles communicated to PME node 8 are more than 2/3 times the 
cut-off out of the domain decomposition cell of their charge group in 
dimension x.
This usually means that your system is not well equilibrated."

I am using 36 cores or a multiple of 36 (i.e. 72, 108, and so on). The 
weird things is that restarting the crashed replica from the checkpoint, 
it starts without problem and it goes for a while (hours). Then it 
crashes again and so on. I have checked the last frame before the crash 
and I do not see structural problem. Changing the number of cores from 1 
to 5 improves the things but not too much, it still crashes and 
restarts. IMHO, if there was a problem with the system the simulation 
should never restart, isn't?

any clue?

Hope I have been clear

Best

Andrea


-- 
Andrea Spitaleri PhD
Computational mOdelling of NanosCalE and bioPhysical sysTems - CONCEPT Lab
ISTITUTO ITALIANO DI TECNOLOGIA
Via Morego 30, 16163 - Genova, Italy
http://it.linkedin.com/in/andreaspitaleri
https://www.researchgate.net/profile/Andrea_Spitaleri
https://iit.it/andrea-spitaleri
cell: +39 3485188790

ORCID: http://orcid.org/0000-0003-3012-3557