normal Simulation crashes well into production

  • shaharsu
  • shaharsu's Avatar Topic Author
  • Visitor
12 years 2 months ago - 12 years 2 months ago #913 by shaharsu
Simulation crashes well into production was created by shaharsu
Hi everyone,

I'm running a simulation box containing a peptide, water and a cosolute modelled according to the polymer martini extension. Minimization is fine and the production starts without any problems (ts = 0.025).

Some time into the simulation, the system crashes with the message:

imb F 1% step 3790400, will finish Wed Jan 18 15:45:49 2012
rank 4 in job 26 shrek14_59024 caused collective abort of all ranks
exit status of rank 4: killed by signal 11

An examination of the log file shows that everything is fine:

Step Time Lambda
3790000 94750.00000 0.00000

Energies (kJ/mol)
Bond Angle G96Angle Proper Dih. Improper Dih.
3.39509e+04 1.85205e+04 2.23299e+01 1.55475e+04 1.02875e+00
LJ (SR) Coulomb (SR) Potential Kinetic En. Total Energy
-1.62958e+06 1.24952e+01 -1.56153e+06 2.51858e+05 -1.30967e+06
Temperature Pressure (bar) Cons. rmsd ()
2.99610e+02 1.66887e-01 3.46313e-06

In addition, and what puzzles me the most, is that I can actually restart the system from the checkpoint, but ONLY if I change the checkpointing interval (-cpt option) to a different number. This lets me run the simulation for an addition length of time (anything between 100k to 1million steps) before crashing again.

The process can be repeated indefinitely, and as far as the trajectory and energies look - everything seems fine.

Any ideas?

Thanks!

Shahar.

Please Log in or Create an account to join the conversation.

  • shaharsu
  • shaharsu's Avatar Topic Author
  • Visitor
12 years 2 months ago #917 by shaharsu
Replied by shaharsu on topic Simulation crashes well into production
I think I've found the problem - when I remove the dihedral potentials of the polymer the simulations run without a problem - must be something wrong there. Moving the discussion to another forum....

** UPDATE **

Anyone who has encountered this problem - reducing the timestep worked and gave stable productions. I'm not sure if this is always the case when using dihedrals, but for the polymer potentials (PEO) I used a timestep of 10 fs which was fine. Faster than that and I had constant crashes.

Shahar.

Please Log in or Create an account to join the conversation.

More
11 years 10 months ago #941 by xavier
Replied by xavier on topic Simulation crashes well into production

shaharsu wrote: I think I've found the problem - when I remove the dihedral potentials of the polymer the simulations run without a problem - must be something wrong there. Moving the discussion to another forum....

** UPDATE **

Anyone who has encountered this problem - reducing the timestep worked and gave stable productions. I'm not sure if this is always the case when using dihedrals, but for the polymer potentials (PEO) I used a timestep of 10 fs which was fine. Faster than that and I had constant crashes.

Shahar.


Hi, sorry for the delay. We have had trouble with the server and not been informed that the forum was active.

The problem you report is known. The dihedral are extremely unstable. The reason is that when three bead defining a dihedral align the dihedral can not be defined and at proximity of this value the system becomes very unstable and it crashes.

It is recommended not to use dihedrals for the backbone but to use "local network"

XAvier.

Please Log in or Create an account to join the conversation.

Time to create page: 0.091 seconds