I am using a server to run MD simulation of my protein. According to instruction of the server it is showing that "Use qsub to run your jobs else your jobs will be automatically killed after sometime."
As I have done direct jobs throughout my MD simulation work, but I am facing this problem with the new server. How to prepare the script file (example) of qsub?
The main difference between running a (MD) job interactively or via an queuing system is that you have to set up a job script file. Essentially, such a file is a shell script and contains all commands to run your application, i.e. invocation of gromacs with the correct command line options. In addition, there are some additional settings for the queuing system, that either may be given as command line options to the qsub command or written into the job script file.
Queuing system options within the job script file depend on the type of queing system used. For e.g. torque, such options start with #PBS (no space!) making them a simple comment in pure shell context.
I strongly recommend to read the man page of qsub for this (or google some first additional infos)!
A general job script (run.pbs) for the bash would be:
#/bin/bash -l
# First the the name
#PBS -N Name-of-job
# How much time to allocate for the run hh:mm:ss
#PBS -l walltime=10:00:00
# Now how many compute nodes and processors should be allocated
# (consult the specific cluster documentation or ask your admin)
#PBS -l nodes=1:ppn=16
# Now we have to change to the directory, the 'qsub' was issued from
# to find the input files and write the output files correctly.
# The environment variable $PBS_O_WORKDIR contains that info.
cd $PBS_O_WORKDIR
echo -n We start our calculation:
date
# Run Gromacs, Amber, Charm, Gaussian, Gamess, Mopac, Vamp ...
echo Done with our calculation:
date
# End
This script is the submitted to the queue simply via
qsub run.pbs
Hope this helps!
Regards,
Anselm
The main difference between running a (MD) job interactively or via an queuing system is that you have to set up a job script file. Essentially, such a file is a shell script and contains all commands to run your application, i.e. invocation of gromacs with the correct command line options. In addition, there are some additional settings for the queuing system, that either may be given as command line options to the qsub command or written into the job script file.
Queuing system options within the job script file depend on the type of queing system used. For e.g. torque, such options start with #PBS (no space!) making them a simple comment in pure shell context.
I strongly recommend to read the man page of qsub for this (or google some first additional infos)!
A general job script (run.pbs) for the bash would be:
#/bin/bash -l
# First the the name
#PBS -N Name-of-job
# How much time to allocate for the run hh:mm:ss
#PBS -l walltime=10:00:00
# Now how many compute nodes and processors should be allocated
# (consult the specific cluster documentation or ask your admin)
#PBS -l nodes=1:ppn=16
# Now we have to change to the directory, the 'qsub' was issued from
# to find the input files and write the output files correctly.
# The environment variable $PBS_O_WORKDIR contains that info.
cd $PBS_O_WORKDIR
echo -n We start our calculation:
date
# Run Gromacs, Amber, Charm, Gaussian, Gamess, Mopac, Vamp ...
echo Done with our calculation:
date
# End
This script is the submitted to the queue simply via
qsub run.pbs
Hope this helps!
Regards,
Anselm
In general - you should find out, which queuing system is running on cluster and what manuals should be consulted when creating job file. The best way is to consult technical team of your cluster to provide you with some working samples and local policy descptions. The example above (from Anselm Horn) is for torque or may be some other pbs-like environment. If it's your case - you may just use that sample - it's rather usefull. Have a look at
man pbs_resources - it should contain data on what you can ask for in "-l".
Notice your ulimit settings - it may change between interactive and batch session. Use ulimit to adjust if you have memory or stack-related issues.
Remember to always CD into the correct directory in the very begining of your script - it will got executed in some not-so-well-predictable place!
It's handy (if implemented) to make torque send you e-mails about you job (-m and -M options of qsub - see manual).
You usually can see your execution queue with qstat command.
#!/bin/bash
#PBS -S /bin/bash
#PBS -N nvt_eq
#PBS -q workq
#PBS -l nodes=2:ppn=40
#PBS -j oe
cd $PBS_O_WORKDIR
module load gromacs514-openmpiintel
echo "The nodefile for this job is stored at ${PBS_NODEFILE}"
cat ${PBS_NODEFILE}
np=$(wc -l < ${PBS_NODEFILE})
echo "Cores assigned: ${np}"
mpirun -np $np gmx_mpi mdrun -v -s ~/directory/nvt.tpr -deffnm nvt
gmx_mpi grompp -f npt.mdp -c nvt.gro -t nvt.cpt -p topol.top -o npt.tpr
mpirun -np $np gmx_mpi mdrun -v -s npt.tpr
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Bioinformatics
- Bioinformatic Tools