AMBER Archive (2008)

Subject: RE: AMBER: Restarting Problem in MD Simulation

From: Ross Walker (ross_at_rosswalker.co.uk)
Date: Fri Jul 18 2008 - 12:08:46 CDT

Hi Khuram,

 

This sounds to me like your restart file may have been corrupted. Start by
checking the first few lines of the restart file - see what it lists as the
time at the top of this file. This is the time from which you would be
restarting. It will likely be earlier than the last entry in the output file
since restrt files are not written on every step but every ntwr steps for
performance reasons. Corruption with restart, mdcrd and mdout files is why I
always suggest running large jobs in pieces. I.e. you can set yourself up a
loop to run in say 500ps chunks (or smaller / larger depending on the
simulation, I generally aim for around 6 to 8 hours per chunk so at most you
would lose 8 hours computation). Thus you'd end up with sequentially
numbered files. In this way if your simulation failed at say 2.12ns and you
found the restart file at 2.10ns was corrupted then you could just go back
to the restart file you have saved for 2.0ns and you don't lose much.

 

In your case if the restart header looks good check through it and see if
there look to be any corruption issues, particularly at the end of the file
if your machine crashed while the restart was being written. You could also
see if ptraj will read the restart file and produce you a pdb file.

 

You should also check through and see if there are any *'s in your restart
file. This can happen with periodic simulations after around 50ns or so
because water molecules can diffuse a long way from the central box and end
up with coordinates that exceed the space in the restart file. A solution is
to use iwrap=1 during the simulation or go back to the previous good restart
file and have ptraj wrap it for you.

 

In your case, however, it is non-periodic boundaries and so there is not the
option to wrap. You are using the Berendsen thermostat with GB which is not
a great idea these days (ntt=3 is generally considered better) and this
thermostat can end up generating center of mass translation which will
ultimately lead to your molecule diffusing a long distance if you don't have
it removed periodicaly with nscm=1000 say. You don't manually change nscm
from the default though so I doubt this is a problem. You may want to check
for *'s in the restart file though just in case, perhaps you had two ligands
in your simulation that have drifted away from each other in different
directions.

 

In summary though if your restart file is corrupted there really isn't much
you can do except go back to the previous good restart file that you have
which in your case sounds like it is the beginning of the simulation.

 

Sorry I can't help much more.

 

Ross

 

 

From: owner-amber_at_scripps.edu [mailto:owner-amber_at_scripps.edu] On Behalf Of
khuram sb
Sent: Friday, July 18, 2008 3:37 AM
To: amber_at_scripps.edu
Subject: AMBER: Restarting Problem in MD Simulation

 

Hi,
I am running MD Simulation. I have done the minimization and equilibration
steps.
When I was running MD Simulation step, my Computer was restarted mistakenly.
My job was performed till 1130ps (742000 steps) out of 1500ps (not in
script form).
Form this MD Simulation the following rst file for final velocity
coordinates was generated:
b_md.rst

I was restarting it using a different named output, rst and mdcrd files,
I am giving the -c b_md.rst as an initial coordinats for the rest of the MD
Simulation process.
My input is:

1IV5 :MD2 Simulation
&cntrl
  imin = 0,
  irest = 1,
  ntx = 5,
  ntb = 0,
  pres0 = 1,
  ntp = 0,
  taup = 1,
  cut = 12,
  ntc = 1,
  ntf = 1,
  igb = 1,
  ntr = 0,
  tempi = 300.0,
  temp0 = 300.0,
  ntt = 1,
  nstlim = 258000,
  dt = 0.0015,
  ntpr = 250,
  ntwx = 250,
  ntwr = 5000
 /

 but it is showing the following error:

______________________________

________________________________________
zuh_at_mcg18:~/Working/1IV5> forrtl: severe (64): input conversion error, unit
9, file /home/zuh/Working/1IV5/b_md.rst
Image PC Routine Line Source
sander 084940D7 Unknown Unknown Unknown
sander 084936F7 Unknown Unknown Unknown
sander 08454C1A Unknown Unknown Unknown
sander 08417CF8 Unknown Unknown Unknown
sander 08417993 Unknown Unknown Unknown
sander 08430B81 Unknown Unknown Unknown
sander 080EF56D Unknown Unknown Unknown
sander 080C11DA Unknown Unknown Unknown
sander 080BFF05 Unknown Unknown Unknown
sander 0804A1C1 Unknown Unknown Unknown
libc.so.6 B7DA9F9C Unknown Unknown Unknown
sander 0804A101 Unknown Unknown Unknown
__________________________________________________________

Can it be restarted from 1130ps again?
If yes than what changes should I make in my input file.
 
My first input file for 1500ps is given below:

1IV5 :MD2 Simulation for 1500ps
 &cntrl
  imin = 0,
  irest = 1,
  ntx = 5,
  ntb = 0,
  pres0 = 1,
  ntp = 0,
  taup = 1,
  cut = 12,
  ntc = 1,
  ntf = 1,
  igb = 1,
  ntr = 0,
  tempi = 300.0,
  temp0 = 300.0,
  ntt = 1,
  nstlim = 1000000,
  dt = 0.0015,
  ntpr = 250,
  ntwx = 250,
  ntwr = 5000
 /

This job took almost 5 days.

Thanks
Khuram SB

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
      to majordomo_at_scripps.edu