AW: Reading files for multiple threaded simulations

From: Norman Geist (norman.geist_at_uni-greifswald.de)
Date: Fri Nov 04 2011 - 02:28:50 CDT

You should place the file where the other namd inputs are psf, pdb (top,
crd) etc, because these files have to be accessible for all nodes, too.

 

Also you said you ran the job on single node. Try to test the job on all
nodes as single node run, maybe it's only a problem of one node. I guess the
nodes have multiple cores, so if you run on single node, but with multiple
cores, you have already ran a parallel job that worked, then it could really
be a problem of the file organization, because there's in case not much
difference between running on one node with multiple processes or on
multiple nodes with multiple processes.

 

Good luck

 

Norman Geist.

 

Von: Yi He [mailto:yi.he.ustcer_at_gmail.com]
Gesendet: Donnerstag, 3. November 2011 16:57
An: Norman Geist
Betreff: Re: namd-l: Reading files for multiple threaded simulations

 

Thanks a lot for the help. I am not sure how the files are organized on the
cluster. It is a school cluster so I don't have the access to nodes other
than the head node. Neither do I have the node definition file used for
mpirun. I think I need to ask the technical support of the cluster for help
on that. I guess I know what is going on now. That file I am trying to read
is generated in the inp script so perhaps it is not saved in the correct
location on the computing nodes. Probably I should generate it before the
namd run.

Best wishes,

Yi

On Thu, Nov 3, 2011 at 3:26 AM, Norman Geist
<norman.geist_at_uni-greifswald.de> wrote:

Hi,

 

Looks that the file is in your local home directory. All nodes you want to
run must have the same file under the same path, are you sure that every
node has your home directory or how do you manage that?? Try to use a volume
that can be reached by all nodes, something like a nfs share. You can check
this by logging on to the nodes and try to access the file.

 

Let me know

 

Norman Geist.

 

Von: owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu] Im Auftrag
von Yi He

Gesendet: Mittwoch, 2. November 2011 17:00
An: Chris Harrison
Cc: Norman Geist; Namd Mailing List

Betreff: Re: namd-l: Reading files for multiple threaded simulations

 

Thanks so much for the reply.
Yes, in the log file, the error I got is when namd2 was trying to read the
file step6.1_equilibration.dihe.
Info: READING EXTRA BONDS FILE
/home/he.yi/simulation/explict/dmpc/equlibrate/flat/namd/restraints/step6.1_
equilibration.dihe
FATAL ERROR: BUFFER OVERRUN - buffer=**
Which is fine when I looked at this file.
Also this error will not appear if I ran namd2 using single node.

Bests,

Yi

On Wed, Nov 2, 2011 at 2:57 AM, Chris Harrison <charris5_at_gmail.com> wrote:

Also, send the log file output. It's rather hard to diagnose with just
"BUFFER OVERRUN." If there's no log file output, then send the job
scheduler output and error file, if any.

Best,
Chris

--
Chris Harrison, Ph.D.
Theoretical and Computational Biophysics Group
NIH Resource for Macromolecular Modeling and Bioinformatics
Beckman Institute for Advanced Science and Technology
University of Illinois, 405 N. Mathews Ave., Urbana, IL 61801
char_at_ks.uiuc.edu                          Voice: 773-570-0329
http://www.ks.uiuc.edu/~char <http://www.ks.uiuc.edu/%7Echar>
Fax:   217-244-6078
Norman Geist <norman.geist_at_uni-greifswald.de> writes:
> Date: Wed, 2 Nov 2011 07:45:32 +0100
> From: Norman Geist <norman.geist_at_uni-greifswald.de>
> To: 'Yi He' <yi.he.ustcer_at_gmail.com>
> Cc: Namd Mailing List <namd-l_at_ks.uiuc.edu>
> Subject: AW: namd-l: Reading files for multiple threaded simulations
> X-Mailer: Microsoft Office Outlook 12.0
>
> Hi
>
>
>
> All mpi processes should have access to the same files under the same
path,
> this is usual for any mpi application. I never observed such problems with
> mpi processes accessing the same files when they are there for everyone.
Try
> to copy your stuff directly into the script and try again to see if it is
> not a problem of your config.
>
>
>
> Let me know
>
>
>
> Norman Geist.
>
>
>
> Von: owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu] Im Auftrag
> von Yi He
> Gesendet: Dienstag, 1. November 2011 18:17
> An: namd-l_at_ks.uiuc.edu
> Betreff: namd-l: Reading files for multiple threaded simulations
>
>
>
> Dear all,
> This may sounds newbie but I want to ask how to avoid errors when you let
> the script read a file (say restraints etc.) in a multiple node run.
> I had a membrane simulation that I need to constraint the head groups at a
> certain position for the whole procedure of the simulation.
> I put something like in the input file:
>
> colvarsConfig restraints/$outputname.col
>
> If I run this as a single node, the simulation runs without a problem.
> If I run this using MPI on the cluster
> mpirun -np 64  -machinefile $PBS_NODEFILE
> /home/he.yi/simulation/softwares/NAMD_2.8_Linux-x86_64/namd2 eql.inp >
> eql.out
> then the simulation has a error like this:
> BUFFER OVERRUN
> This happens when namd is trying to read the restraint file under the same
> folder.
> I also have problem if I want to source a file contains parameters.
>
> Yi
>
 
 

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:57:53 CST