QUEUES for CARLES BO group: Difference between revisions

From Wiki
Jump to navigation Jump to search
New page: == SENDING CALCULATIONS: == qs TYPE_OF_CALCULATION inputfile.in where TYPE_OF_CALCULATION can be: g09.c8m24 (8 cores smp pe) g09.cq4m4 (4 cores smp pe) g09.c4m8 (4 co...
 
No edit summary
 
(7 intermediate revisions by one other user not shown)
Line 1: Line 1:
go back to [[Main_Page]], [[Computational Resources]], [[Clusters]], [[Local Clusters]], [[kimik2]]
== SENDING CALCULATIONS: ==
== SENDING CALCULATIONS: ==
       
qs TYPE_OF_CALCULATION inputfile.in


where TYPE_OF_CALCULATION can be:
Until a 'qs' wrapper will be made you must to conform a .sub script manually and send it using 'qsub' command.
 
Below these lines you have an example of .sub script where you must substitute the bold variables:
 
----
''''$JOBNAME''''    identify the job in the cluster queue.
 
''''$QUEUE_NAME''''  identify the queue where you are sending the calculation
* qconf -sql
 
''''$PE_NAME'''''    identify the name of the Parallel Environment
* qconf -sq $QUEUE_NAME | grep pe_list
 
''''$N_PROC''''      identify the requested number of processors
 
''''$ADFVERSION''''  identify the version of ADF that you can use
* adf2007.01 | adf2008.01 | adf2009.01
''''$CALCULATION'''' paste your calculation script
----
     


g09.c8m24 (8 cores smp pe)
g09.cq4m4 (4 cores smp pe)     
g09.c4m8 (4 cores smp pe)


g03.cq4m4 (4 cores smp pe)
g03.c4m8 (4 cores smp pe)


'''NOTE:''' remember to add the correct parameters in your inputfile.in according to the TYPE_OF_CALCULATION. '''%nproc=...''' and '''%mem=...'''
#!/bin/bash                                                       
You can run 8 jobs at the same time.
#$ -N '''$JOBNAME'''                                                   
#$ -cwd                                                           
#$ -masterq '''$QUEUE_NAME'''                                           
#$ -pe '''$PE_NAME $N_PROC'''
#Set an environment variable for the location of MPI
#$ -v MPIR_HOME=/usr/local/adf2008.01/bin/hpmpi   
                                                   
#The number of slots (CPUs) grid engine allocates 
#is set as the variable $NSLOTS                   
echo "Got $NSLOTS slots."                         
export PATH=$PATH                                             
                                                             
export TMPDIR=/scratch                                       
export ADFHOME=/usr/local/'''$ADFVERSION'''                          
export ADFBIN=$ADFHOME/bin
export ADFRESOURCES=$ADFHOME/atomicdata
export SCMLICENSE=$ADFHOME/license
export SCM_USETMPDIR=yes
export SCM_NODOMAINCHECK=yes
export MPIDIR=$ADFHOME/bin/hpmpi
export PATH=$PATH:$MPIDIR/bin:$TMPDIR
export SCMWISH=""
$ADFBIN/adf << eor
'''$CALCULATION'''
echo QUEUE $QUEUE
echo PE $PE
echo NSCM $NSCM
  echo $PWD
cd /scratch
ls -lR
rm -rf *

Latest revision as of 10:03, 21 September 2011

go back to Main_Page, Computational Resources, Clusters, Local Clusters, kimik2

SENDING CALCULATIONS:[edit]

Until a 'qs' wrapper will be made you must to conform a .sub script manually and send it using 'qsub' command.

Below these lines you have an example of .sub script where you must substitute the bold variables:


'$JOBNAME' identify the job in the cluster queue.

'$QUEUE_NAME' identify the queue where you are sending the calculation

  • qconf -sql

'$PE_NAME identify the name of the Parallel Environment

  • qconf -sq $QUEUE_NAME | grep pe_list

'$N_PROC' identify the requested number of processors

'$ADFVERSION' identify the version of ADF that you can use

  • adf2007.01 | adf2008.01 | adf2009.01

'$CALCULATION' paste your calculation script




#!/bin/bash                                                        
#$ -N $JOBNAME                                                     
#$ -cwd                                                            
#$ -masterq $QUEUE_NAME                                             
#$ -pe $PE_NAME $N_PROC

#Set an environment variable for the location of MPI
#$ -v MPIR_HOME=/usr/local/adf2008.01/bin/hpmpi     
                                                   
#The number of slots (CPUs) grid engine allocates   
#is set as the variable $NSLOTS                     
echo "Got $NSLOTS slots."                           

export PATH=$PATH                                              
                                                              
export TMPDIR=/scratch                                         
export ADFHOME=/usr/local/$ADFVERSION                           
export ADFBIN=$ADFHOME/bin
export ADFRESOURCES=$ADFHOME/atomicdata
export SCMLICENSE=$ADFHOME/license
export SCM_USETMPDIR=yes
export SCM_NODOMAINCHECK=yes
export MPIDIR=$ADFHOME/bin/hpmpi
export PATH=$PATH:$MPIDIR/bin:$TMPDIR
export SCMWISH=""

$ADFBIN/adf << eor

$CALCULATION

echo QUEUE $QUEUE
echo PE $PE
echo NSCM $NSCM
 echo $PWD
cd /scratch
ls -lR
rm -rf *