Dear Dongshan,
The problem is that the scratch directory is defined in the rungms
script; and the scratch directory on a cluster has to be potentially
unique for each job (to avoid to cross temporary files among different
jobs); so we decided to generate a unique rungms by PBS job...
- First we created a generic 'rungms.RED-Server' script that contains
a generic ZZZZZZ scratch directory:
The PBS script will replace ZZZZZZ by a specific name using its
internal variable ($PBS_JOBID):
# set SCR=/scr/$USER
# set USERSCR=~$USER/scr
# set GMSPATH=/u1/mike/gamess
set SCR=/tmp/ZZZZZZ
set USERSCR=/tmp/ZZZZZZ
set GMSPATH=/usr/local/gamess
- Then R.E.D. III.52 can automatically search for a 'rungms.info' file
(see below); 'rungms.info' contains the name of the rungms$X file to
be called by R.E.D...:
# To use R.E.D./GAMESS on a UNIX cluster using a PBS script to handle
multiple copies of the rungms$X script
if (-e "$pathgamessus/rungms.info"){
open(INFO,"< $pathgamessus/rungms.info");
foreach(<INFO>){ $temprungms=$_; }
close(INFO);
chomp($temprungms);
chomp($rungms=`which $temprungms`);
}
else { chomp($rungms=`which rungms`); }
- Finally attached a part of our PBS script (csh shell) used on R.E.D.
Server...
if you find a more easy way - I am very much interested in ;-)
regards, Francois
> I used the following PBS script to submit a job on the cluster to do the
> geometry optimization and create the .log file:
>
> ==================================================
> #!/bin/bash
> #PBS -q tiny16core
> #PBS -l walltime=2:00:00
> #PBS -l nodes=1:ppn=16
> #PBS -N gms-16
>
> echo $PBS_O_WORKDIR
> #module load intel/12.1.5
> #module load mkl/12.1.5
> #module load mvapich2/1.8
> #module load gamess/5.1.2012
>
> cd $PBS_O_WORKDIR
> #cat $PBS_NODEFILE
> #NP=$(wc -l < $PBS_NODEFILE)
>
> /share/apps/gamess/rungms mel-gam 01 1 16 > mel-gam.log
>
> ===============================================
>
> I can finish this step and get the .log file normally.
>
> While for the RED charge fitting, I did through a PBS job in a queuing
> system or through an interactive job on the login node. but both of these
> ways failed. For the interactive running on the login node, I used "perl
> RED-vIII.52.pl" and for the pbs script, the last line in the above file
> changed to "perl RED-vIII.52.pl".
>
> Best!
>
> Dongshan
>
>
> On Wed, Oct 9, 2013 at 3:03 PM, FyD <fyd.q4md-forcefieldtools.org> wrote:
>
>> Dear Dongshan,
>>
>> > I gave up to use the cluster to execute RED charge fitting and changed to
>> > use the RED server on your website. Now I can finish the charge fitting
>> > task and get the .mol2 file for Amber using. I notice that the RED server
>> > use RED-III.4 instead of RED-III.52 to execute the charge fitting
>> > procedure. Is there really bugs in RED-III.52 tools?
>>
>> R.E.D. Server interfaces R.E.D. IV, while the R.E.D. III.x tools are
>> distributed under the GPL license.
>>
>> I can help you to configure your cluster and send you our scripts...
>>
>> You did not answer to the following question:
>> Do you use a queuing system on your cluster? if Yes do you use PBS?
>>
>> regards,
>> Francois
>>
>>
>> > On Tue, Oct 8, 2013 at 11:37 PM, Dongshan Wei <dswei0523.gmail.com>
>> wrote:
>> >
>> >> Dear Francois,
>> >>
>> >> I'm still having questions about the execution of RED-III.52. When I
>> >> execute the command "perl RED-vIII.5.pl", I got the same question with
>> >> the following link:
>> >> http://q4md-forcefieldtools.org/RED/popup/bugs-III.4.txt
>> >>
>> >> "Problem: The scratch path defined in the "rungms" script of the
>> GAMESS-US
>> >> program is not found by R.E.D.-III.4 if an environment variable such as
>> >> $USER is used in the path. For instance the /tmp/$USER path is not found
>> >> even if the corresponding directory does exist (environment $USER
>> variable
>> >> points to "bill"), while /tmp/bill is found. This leads to the following
>> >> error message: "
>> >>
>> >> So RED-III.52 doesn't update the bug? I want to go back to use
>> RED-III.4,
>> >> but it seems the RESP web doesn't offer the download for RESP-III.4
>> tools.
>> >> Is it true?
>> >>
>> >> Best!
>> >>
>> >> Dongshan
>> >>
>> >>
>> >> On Tue, Oct 1, 2013 at 1:35 PM, FyD <fyd.q4md-forcefieldtools.org>
>> wrote:
>> >>
>> >>> Dear Dongshan,
>> >>>
>> >>> Do you use the PBS queuing system?
>> >>>
>> >>> At R.E.D. Server we do use PBS (Torque/Maui):
>> >>> See http://q4md-forcefieldtools.org/REDS/faq.php#8
>> >>> We also use the same gamess version...
>> >>>
>> >>> # set SCR=/scr/$USER
>> >>> # set USERSCR=~$USER/scr
>> >>> # set GMSPATH=/u1/mike/gamess
>> >>> set SCR=/tmp/ZZZZZZ
>> >>> set USERSCR=/tmp/ZZZZZZ
>> >>> set GMSPATH=/usr/local/gamess
>> >>>
>> >>> We use the same direction for SCR and GMSPATH; R.E.D. will move the
>> >>> punch file from the scratch directory to the working directory to
>> >>> extract the MEP...
>> >>>
>> >>> and the PBS script modify rungms to create a unique & temporary
>> >>> scratch directory by job:
>> >>> see and example:
>> >>> set SCR=/tmp/R.E.D.IV-4379.master0.q4md-forcefieldtools.org
>> >>> set USERSCR=/tmp/R.E.D.IV-4379.master0.q4md-forcefieldtools.org
>> >>> set GMSPATH=/usr/local/gamess
>> >>>
>> >>> ---
>> >>>
>> >>> the definition of the punch file is done in the script gms-files.csh
>> >>> setenv INPUT $SCR/$JOB.F05
>> >>> setenv PUNCH $USERSCR/$JOB.dat
>> >>> setenv AOINTS $SCR/$JOB.F08
>> >>> setenv MOINTS $SCR/$JOB.F09
>> >>> setenv DICTNRY $SCR/$JOB.F10
>> >>>
>> >>> regards, Francois
>> >>>
>> >>>
>> >>> > Thanks for your detailed reply. The gamess I used on the cluster is
>> >>> gamess
>> >>> > 5.1.2012 version. The rungms script is original from the gamess
>> software
>> >>> > and is attached here. Please help me have a look.
>> >>> >
>> >>> > Especially, the scratch directory /local_scratch/dswei disappeares
>> after
>> >>> > the job running finishes. I tried to change the scratch directory to
>> my
>> >>> > user directory in the rungms script and submitted a job to the
>> cluster,
>> >>> but
>> >>> > the job quit quickly since there is no comminucation between the node
>> >>> and
>> >>> > my user directory. So how can I create a scratch directory both for
>> job
>> >>> > running and output saving?
>> >>>
>> >>> > On Tue, Oct 1, 2013 at 4:07 AM, FyD <fyd.q4md-forcefieldtools.org>
>> >>> wrote:
>> >>> >
>> >>> >> Dear Dongshan,
>> >>> >>
>> >>> >> > Did you meet one problem "ERROR: Problem with the QM Scratch
>> >>> directory"
>> >>> >> > when using R.E.D. VIII.52 to derive the charge of atoms? In my
>> case,
>> >>> I
>> >>> >> had
>> >>> >> > used gamess to optimized the molecular structure and when I
>> executed
>> >>> >> R.E.D.
>> >>> >> > VIII.52 with "perl RED-vIII.5.pl" command, the error is always
>> >>> showing
>> >>> >> as
>> >>> >> > follows:
>> >>> >>
>> >>> >> which GAMESS do you interface?
>> >>> >>
>> >>> >> > The Scratch directory defined for GAMESS is /local_scratch/$USER
>> >>> >> >
>> >>> >> > Scratch directory for GAMESS [ NOT
>> >>> FOUND ]
>> >>> >>
>> >>> >> So you need to create the corresponding scratch directory before to
>> >>> >> execute R.E.D.
>> >>> >>
>> >>> >> > The USER Scratch directory defined for GAMESS is
>> >>> >> /scratch/$USER
>> >>> >> > Bad permissions for the GAMESS Scratch directory
>> >>> >>
>> >>> >> 'chmod 777' for this scratch directory...
>> >>> >>
>> >>> >> > The punch file directory defined for GAMESS is
>> >>> >> $USERSCR/$JOB.dat
>> >>> >> >
>> >>> >> > ERROR: Problem with the QM Scratch directory.
>> >>> >> > See the R.E.D.-II manual, page 7.
>> >>> >>
>> >>> >> See http://q4md-forcefieldtools.org/RED/RED-II.pdf
>> >>> >>
>> >>> >> > About the above error, I have some questions to ask:
>> >>> >> > (1) To execute R.E.D.VIII.52, how many input files need to be
>> read? I
>> >>> >> think
>> >>> >> > the necessary files are only three: Mol-red1.p2n, Mol-red1.log and
>> >>> >> > Mol-red1.dat. Is this correct?
>> >>> >>
>> >>> >> See http://q4md-forcefieldtools.org/Tutorial/Tutorial-1.php#7
>> >>> >> - if $OPT_Calc = "On" & $MEPCHR_Calc = "On" only the P2N file(s)
>> >>> >> is(are) required; i.e. in your case Mol_red1.p2n
>> >>> >> - if $OPT_Calc = "Off" & $MEPCHR_Calc = "On" the P2N file(s) and QM
>> >>> >> output log file(s) are required; i.e. in your case Mol_red1.p2n &
>> >>> >> Mol_red1.log
>> >>> >> I do not know what is this Mol_red1.dat file; I think you mix up the
>> >>> >> punch file and that Mol_red1.dat file.
>> >>> >> See http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2918240/
>> >>> >> & http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2918240/figure/F1/
>> >>> >> the punch file is the JOB2-gam_m$n-$i-$j.dat file
>> >>> >> $n molecule // $i conformation // $j orientation
>> >>> >>
>> >>> >> > (2) To my understanding, the scratch directory
>> /local_scratch/$USER
>> >>> is
>> >>> >> > generated temporarily for intermediate outputs during the
>> >>> optimization
>> >>> >> > calculation before executing R.E.D. fitting, and after the job is
>> >>> >> finished,
>> >>> >> > the temporal scratch directory will be deleted automaticly. So to
>> do
>> >>> >> R.E.D.
>> >>> >> > fitting, the scratch directory is not necessary. Is this correct?
>> >>> >>
>> >>> >> on a cluster - & using the PBS queuing system, Yes, a temporary
>> >>> >> scratch directory can be created just before executing R.E.D. and
>> >>> >> removed just after the job has finished.
>> >>> >>
>> >>> >> on a workstation you could create this scratch directory only once
>> and
>> >>> >> keep it for all the R.E.D. jobs... If this scratch directory is only
>> >>> >> for your login you could use a specific directory in your login (if
>> >>> >> you have a enough space in your login also)
>> >>> >> If the scratch directory is to be shared with other users you could
>> >>> >> use the /tmp partition and then create a directory in /tmp related
>> to
>> >>> >> your login. /tmp could be a particular partition with a looot of
>> room.
>> >>> >>
>> >>> >> > (3) From the above information, it seems the R.E.D. VIII.52 cannot
>> >>> >> > recognize variables such as $USER, $USERSCR, $JOB, because the
>> >>> directory
>> >>> >> > /scratch/dswei and the file /scratch/dswei/Mol-red1.dat do exist.
>> >>> Here
>> >>> >> > dswei is my username on the cluster. How to solve the problem?
>> >>> >>
>> >>> >> please tell me which GAMESS version you use? and send me your
>> 'rungms'
>> >>> >> script: I could prepare it for you if you agree...
>> >>> >>
>> >>> >> > I'm learning the R.E.D., so your reply to my above question will
>> be
>> >>> great
>> >>> >> > help to me.
>> >>> >>
>> >>> >> do not hesitate to ask as many questions as you wish ;-)
>> >>> >>
>> >>> >> regards, Francois
>> >>>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Oct 10 2013 - 01:30:04 PDT
