Dear Mike,
thanks for the detailed advice.
I did check out the density and it is very stable at 1.05 throughout the
simulation, there are changes only at third decimal place. Also i had a
very close look (visually) at the pdb files of last 10ps of 5ns runs,
generated using ptraj, and they do not seem to have any vacuum bubbles.
Secondly it does run with single processor as you predicted so i suppose
it might be a problem with the biological system and not the
software/hardware. But in that case, i am absolutely puzzled. To tell you
why, i will elaborate what i had done earlier:
After a rigorous minimization protocol involving successively reduced
restraints on the solute and so on (final minimization without
restraints), i heated the system to 300K, then ran a 200ps equilibration.
Using the restart file of equilibration run, i ran 6 production runs upto
5 ns without any problem. Only when i analysed those results and tried to
extend my simulations, i ran into the problem of 'lastrst'. (and i am not
able to extend any of the 6 simulations.) So today just to crosscheck, i
tried to start a production run 'fresh' from the end of equilibration
stage i.e. by using the restart file of equilibration run and again, i got
the 'lastrst' error. Now if there were vacuum bubbles created right during
equilibration or earlier, how did sander run 6 production runs earlier
without any problem? I also tried to run the equilibration again, from the
heating stage restart file and got the same error.
It looks like i may have to start all over again, but then i have no clue
as to why sander did not detect the vacuum bubbles earlier, assuming they
do exist. Any ideas?
Thanks and Regards,
Arvind
ps: pmemd is not installed on the cluster so it may take a while if it
must be checked with pmemd, as per Robert Duke's advice.
> Dear Arvind,
> This is an unfortunately vague error message, and there is a way to fix
> the code so that this failure does not occur, but I have been somewhat
> reluctant to "fix" the problem since it usually indicates that there is
> a problem with the system, not with the program.
>
> Take a close look at your system and see if it is, in fact,
> inhomogeneous. Does it have vacuous regions? Before looking, use ptraj
> to image the solvent back into the primary unit cell.
> Also look at the density reported by sander to determine if it is
> reasonable (very near 1.0 for most solvated biomolecular systems).
> If it is inhomogeneous, then please let me know if you want to run the
> system this way. Normally people do not want vacuum bubbles in their
> systems, an artifact of the setup procedure coupled with not
> equilibrating at constant pressure long enough to reach a reasonable
> density.
>
> You can test that the problem is not with sander by running on one
> processor. Your results from a few steps will look reasonable, at least
> compared to the last frame of the runs you say are successful. On one
> processor, the inhomogeneous problem does not stop the simulation.
>
> If, in fact, your system is inhomogeneous, AND you do want to run that
> way, I can help to modify sander to run for you on 8 processors.
>
> Otherwise you will need to equilibrate to the correct density/volume
> before running for multiple nanoseconds, and throw away the other results.
>
> Best wishes
> Mike
>
>
> Arvind Marathe wrote:
>> Dear amber users,
>> I am running molecular dynamics simulations on a system of 25544 atoms.
>> I
>> had already finished 5 ns simulation without any problem. After
>> analysing
>> those results, when i tried to continue the simulation on the same
>> cluster
>> (SGI_ALTIX, 8 processors), i am getting the following error in my .out
>> file (please see below) which says 'exceeding lastrst in get_stack'. So
>> after searching the archives, i set the lastrst value to a very large
>> number (1000000000000) in my .in file and tried to run the simulation,
>> but
>> with the same error message. Just to check, i tried to run the
>> simulation
>> with a previous restart file for the same biological system which had
>> run
>> succesfully earlier on the same cluster. But even this effort to repeat
>> the previous run failed. Also note that other users are able to run
>> their
>> amber8 jobs on the same cluster (with other biological systems) without
>> getting this error. Any ideas about what and where the problem could be?
>>
>> Thanks and Regards,
>> Arvind
>>
>>
>> Error in .out file
>> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
>>
>> --------------------------------------------------------------------------------
>> 4. RESULTS
>> --------------------------------------------------------------------------------
>>
>> ---------------------------------------------------
>> APPROXIMATING switch and d/dx switch using CUBIC SPLINE INTERPOLATION
>> using 5000.0 points per unit in tabled values
>> TESTING RELATIVE ERROR over r ranging from 0.0 to cutoff
>> | CHECK switch(x): max rel err = 0.3338E-14 at 2.509280
>> | CHECK d/dx switch(x): max rel err = 0.8261E-11 at 2.768360
>> ---------------------------------------------------
>> | Local SIZE OF NONBOND LIST = 912907
>> | TOTAL SIZE OF NONBOND LIST = 7141631
>> ***** Processor 0
>> ***** System must be very inhomogeneous.
>> ***** Readjusting recip sizes.
>> In this slab, Atoms found: 25544 Allocated: 5108
>>
>> Exceeding lastrst in get_stack
>> lastrst = 799562
>> top_stk= 731432
>> isize = 136232
>> request= 867664
>> Increase lastrst in the &cntrl namelist
>>
>> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
>>
>> -----------------------------------------------------------------------
>> The AMBER Mail Reflector
>> To post, send mail to amber.scripps.edu
>> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
>>
>>
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
>
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Sun Sep 17 2006 - 06:07:12 PDT
