Hi Dean, Sasha -
Well, there are a number of ways this can happen, more-likely operating system related, but it could be there are some new wrinkles in the code of which I am unaware. By any chance, are these problems occurring on Cray XT-series machines? I remember there were issues with how file buffering was set up on the Cray XT-3 at PSU several years ago; I would have to do a bit of digging to reconstruct the details, but how file buffering worked was actually controlled from the user environment, if memory serves, and had potential performance and reliability impacts. Anyway, details about the actual systems in use would be helpful. If memory serves, really the only way we could flush the buffers during a run was an actual close and reopen cycle; I should probably look at the code to see if anything new and interesting in that regard. Anyway, complete details about the runs as well as how the software was built can only help.
Regards - Bob Duke
________________________________________
From: Dean Cuebas [deancuebas.missouristate.edu]
Sent: Saturday, December 24, 2011 1:02 PM
To: AMBER Mailing List
Subject: Re: [AMBER] Restart file for pmemd not showing all information
A quick look at the archives shows that several people have had this
problem.
I have observed this problem as well. Might be worth a look, since
incomplete restart files generated during a run are useless.
Dean
On 12/22/11 3:54 PM, "Sasha Perkins" <slperks49.gmail.com> wrote:
>Ross,
>
>Thanks for the prompt reply. Unfortunately, still after killing the job in
>mid cycle, the whole file is still not being written. Additionally, I
>tried
>running this same input file with sander and the complete file IS written
>at all the cycles I ask it to write the rst file. The issue seems to be
>only present with pmemd. Could there be other reasons as to why the
>complete file isn't being written in any of the other cycles besides the
>last one? Or if it is a buffering issue, are there any ways to manually
>force the complete file to be written?
>
>Thanks in advance,
>
>Sasha Perkins
>Graduate Student, Penn State University
>
>
>On Thu, Dec 22, 2011 at 12:10 PM, Ross Walker <ross.rosswalker.co.uk>
>wrote:
>
>> Hi Sasha,
>>
>> I suspect this is a buffering issue with the file system. Have you tried
>> killing a job mid cycle and seeing if the restart file then gets the
>>rest
>> of
>> the information written to it?
>>
>> What is happening is that your operating system is buffering the
>> information
>> being written to the restart file. As far as pmemd is concerned
>>everything
>> got written correctly and indeed if it were to close and reopen the
>>restart
>> file it would be able to read all the information. Also if your job
>> crashes,
>> since it is the operating system that is buffering it so the remaining
>> restart file information should get flushed. Your best option is to try
>> killing a job 'kill -9 PID' and see what happens to the restart file
>>that
>> is
>> written.
>>
>> If it is still not writing it then we can take a look and see what else
>> might be causing it.
>>
>> All the best
>> Ross
>>
>> > -----Original Message-----
>> > From: Sasha Perkins [mailto:slperks49.gmail.com]
>> > Sent: Thursday, December 22, 2011 8:48 AM
>> > To: amber.ambermd.org
>> > Subject: [AMBER] Restart file for pmemd not showing all information
>> >
>> > Hello,
>> >
>> > This is one of my steps in my simulation. I noticed, however, that
>>each
>> > time the restart file is written during my run, there are about 80
>> > atoms
>> > (my system has a total of 9600 atoms) who's information is not written
>> > in
>> > the restart file, as well as the box information is missing. However,
>> > at
>> > the end of the run, all information is written for the last cycle.
>> > Prior
>> > runs with the same input file but using sander showed all information
>> > in
>> > the restart files every time the file was written.
>> >
>> > Below is my input file which I am running using pmemd in amber 11.
>> > &cntrl
>> > imin = 0, irest=1, ntx=5,
>> > ntb=2, ntp=1, taup=1.0,
>> > pres0=0.98,
>> > cut = 20.0,
>> > ntc=2, ntf=2,
>> > temp0=298.0,
>> > ntt=3, gamma_ln=5,
>> > nstlim= 400000, dt =0.002,
>> > ntpr=100, ntwx= 100, ntwr = 1000
>> > &end
>> >
>> > Any helpful information regarding how to fix this, or if this is
>> > normal,
>> > would be greatly appreciated since I need all information to be
>>written
>> > in
>> > the restart files in case my runs unexpectedly get terminated before
>> > the
>> > maximum number of cycles is reached.
>> >
>> > Thank you,
>> >
>> > Sasha Perkins
>> > Graduate Student, Penn State University
>> > _______________________________________________
>> > AMBER mailing list
>> > AMBER.ambermd.org
>> > http://lists.ambermd.org/mailman/listinfo/amber
>>
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
>_______________________________________________
>AMBER mailing list
>AMBER.ambermd.org
>http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sat Dec 24 2011 - 13:30:04 PST
