Dear Amber Users!
I'm running several test simulations using GPU cluster (to assign each md
job on each node consisted of 1 or 2 gpus).
Below you can find my script according to which each of the job consist of
the sequence md runs (minimization-> several equilibration -> productions).
#!/bin/bash
#OAR -n md_OR5AP2_MULTI_octanol
#OAR -l nodes=1,walltime=700:00:00
#OAR -p visu = 'YES'
#OAR -t visu
#OAR -O result.OR5AP2_MULTI_octanol.log
#OAR -E result.OR5AP2_MULTI_octanol.log
#OAR --notify mail:gleb.novikov.unice.fr
module purge
module load ambermd/cuda+intel-13+mpich2-1.5
ulimit -s unlimited
#cd /workspace/gnovikov/sims/md_OR5AP2_MULTI_octanol
#pmemd.cuda -O -i ./in/minim.in -o minim.out -p ./protein.parm7 -c
./protein.inpcrd -r minim.restrt
#pmemd.cuda -O -i ./in/equil.in -o equil.out -p ./protein.parm7 -c
./minim.restrt -r equil.restrt -x equil.nc -ref ./minim.restrt
pmemd.cuda -O -i ./in/equil2.in -o equil2.out -p ./protein.parm7 -c
equil.restrt -r equil2.restrt -x equil2.nc -ref equil.restrt
#pmemd.cuda -O -i ./in/equil3.in -o equil3.out -p ./protein.parm7 -c
equil2.restrt -r equil3.restrt -x equil3.nc -ref equil2.restrt
#pmemd.cuda -O -i ./in/equil4.in -o equil4.out -p ./protein.parm7 -c
equil3.restrt -r equil4.restrt -x equil4.nc -ref equil3.restrt
mpirun -np 2 pmemd.cuda.MPI -O -i ./in/md.in -o md_OR5AP2_MULTI_octanol.out
-p ./protein.parm7 -c equil4.restrt -r md.restrt -x md.nc
Depending on the system it catch a crash on different steps (some systems
are terminated during equilibration, another during production run etc). My
task is to easily locate source of the crashes for each system (e.g to find
it within some useful log) and to fix it for specified case.
By looking within the folder of any terminated system I have not found
anything about errors in log produced for each run as well as could not
find resulted file which could consist of something
#OAR -E result.OR5AP2_MULTI_octanol.log.
Should I modify my input run script to obtain such useful information (e,g
to tun simulation in some debug mode)?
James
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Nov 04 2014 - 00:30:02 PST
