Resubmit your MD simulations automatically

Based on a GROMACS resubmission script by Dr. Camilo Aponte-Santamaria. These scripts will automatically queue multiple days of GROMACS or oxDNA simulations using Slurm dependencies for simulations that take longer than common 24-hour wall times.

resubmit_gromacs.sh has the following 10 required arguments:

n_iter : How many iterations to run
time_limit : Time per iteration
jobname : Name of job in the Slurm queue
n_nodes : How many nodes to request
resources : Allocations of cpus, gpus, and tasks in the format tasks_per_node-cpus_per_task-tasks_per_core-gpus (e.g. 1-18-1-1 for 1 task, 18 cpus for the task (-ntomp 18), no hyperthreading, 1 gpu).
tpr : path to your simulation .tpr file
deffnm : passed to the mdrun -deffnm parameter
outdir : Path to simulation directory
continuation : no if new simulation, yes if continuing from a .cpt file.  Your .cpt file must be named <deffnm>.cpt
submit : no to create Slurm files but not submit, yes to submit once files are created.

Run resubmit_gromacs.sh -h to print the help to the cli.

• The continuation and submit options must be 'yes' or 'no'.
• If continuing a simulation, your .cpt file must be named <deffnm>.cpt.
• gmx_mpi must be on your $PATH (if using just gmx or you have a path, edit the script).

resubmit_oxDNA.sh has the following 8 required arguments:

n_iter : How many iterations to run
time_limit : Time per iteration
jobname : Name of job in the Slurm queue
input : Path to the oxDNA input file
steps : Total number of steps the simulation should run for
outdir  Path to simulation directory
continuation : no if new simulation, yes if continuing from a lastconf file.  Your lastconf file must be named last_conf.dat
submit : no to create Slurm files but not submit, yes to submit once files are created.

Run resubmit_oxDNA.sh -h to print the help to the cli.

• The continuation and submit options must be 'yes' or 'no'.
• For a continuation, the lastconf file must be in the simulation directory and must be named last_conf.dat.
• Your oxDNA binary must be at ~/software/oxDNA/build/bin/oxDNA (or edit the script).

oxDNA does not use much GPU memory, job throughput can be significantly incrased using Nvidia Multiprocessing Service (see this benchmark). This script will create n_replicates replicates of the specified simulation.

resubmit_oxDNA_mps.sh has the following 9 required arguments:

n_iter : How many iterations to run
time_limit : Time per iteration
jobname : Name of job in the Slurm queue
input : Path to the oxDNA input file
steps : Total number of steps the simulation should run for
n_replicates : Number of jobs to run in parallel
outdir  Path to simulation directory
continuation : no if new simulation, yes if continuing from a lastconf file.  Your lastconf file must be named last_conf.dat
submit : no to create Slurm files but not submit, yes to submit once files are created.

Run resubmit_oxDNA_mps.sh -h to print the help to the cli.

• The continuation and submit options must be 'yes' or 'no'.
• For a continuation, the lastconf file must be in the simulation directory and must be named last_conf.dat.
• Your oxDNA binary must be at ~/software/oxDNA/build/bin/oxDNA (or edit the script).
• Your input, topology and initial configuration should be in the parent directory of the replicate runs.
• This script will create n_replicates directories each called run#. If they already exist, it will still copy the input and all topology and configurations from the current directory into each run directory.