Nothing is written to restart files

LAMMPS LAMMPS Mailing List Mirror
#1

Hello,

I am trying to run a large (~1,000,000 atom) simulation for a long period of time using restarts, but for some reason the data is not written to the restart files. The files are created at the correct intervals, but are only 20 bytes in size. I have included the input file below. I have tried the same restart command with another very similar input file, but a much smaller system, and that seems to work fine. I am using LAMMPS (18 Sep 2010). Thank you,

Wylie

Note: I am writing restarts every 1000 timesteps just as a debugging procedure.

# 3-d DPD flow simulation

dimension 3
boundary p p p
units real
atom_style full
read_data project.data
neighbor 2.0 bin
communicate single vel yes
neigh_modify delay 1
bond_style harmonic
bond_coeff 1 1000.0 7.82 #5.276
angle_style harmonic
angle_coeff 1 6800.0 180

pair_style hybrid/overlay morse 16.0 dpd/tstat 300. 300. 10 44355
pair_coeff * * morse 0.837 0.35 4.7 #dpd/tstat 1.0 #water,boundary and non-polar chains (also try 0.43)
pair_coeff 1 1 morse 0.813 0.5552 6.29 #dpd/tstat 1.0 #water
pair_coeff 1 2 morse 0.8 0.55 6.26 #dpd/tstat 1.0 # water and surface
pair_coeff 1 5 morse 0.8 0.55 6.29 #dpd/tstat 1.0 # water and cover
pair_coeff 2 2 morse 0.8 0.55 3.536 #dpd/tstat 1.0 # surface and surface
pair_coeff 1 3 morse 0.8 0.55 6.29 #dpd/tstat 1.0 # water and cnts
pair_coeff 2 3 morse 0.2 0.55 6.29 #dpd/tstat 1.0 # surface and cnts
pair_coeff 3 3 morse 0.2 0.55 6.29 #dpd/tstat 1.0 # cnts and cnts
pair_coeff 3 5 morse 0.2 0.55 6.29 #dpd/tstat 1.0 # cnts and cover
pair_coeff * * dpd/tstat 2200.0
# Run

timestep 20
thermo_style custom step temp pe etotal press vol
thermo 500
#minimize 0.0 1.0e-12 10000 20000

group boundary type 2 5
group flow subtract all boundary
compute 10 flow msd
fix 73 all ave/time 1 1 1 c_10[4] file msd.data

#### Restart #######
restart 1000 flow.restart

velocity flow create 300.00 376847
fix 44 boundary spring/self 10.0
velocity boundary create 300.0 33234
fix 4 all nve
dump 23 all atom 5000 projectD0.dump
dump 24 all dcd 1000 projectD0.dcd

run 50000

pair_coeff * 5 morse 0.0 1.0 1.0 # Turn off cover
pair_coeff * 5 dpd/tstat 0.0
run 600000

#2

Try adding this line to bench/in.lj

restart 50 tmp.restart

and running it as

lmp_foo -v x 4 -v y 4 -z y 4 < in.lj

which is about 2M atoms.

It works fine on 12 procs for me.

Steve

#3

Hello,

I am trying to run a large (~1,000,000 atom) simulation for a long period of time using restarts, but for some reason the data is not written to the restart files. The files are created at the correct intervals, but are only 20 bytes in size. I have included the input file below. I have tried the same restart command with another very similar input file, but a much smaller system, and that seems to work fine. I am using LAMMPS (18 Sep 2010). Thank you,

this is very difficult to debug, but the first recommendation would be
to upgrade
to the current version of LAMMPS. a lot of effort has been made to handle
large systems well. the input file doesn't really help, if the problem
is problem
size related. why don't you try using the lj benchmark input with proper options
to replicate and a restart command added as a test case and see, when
exactly you run into problems and see, if that persists in the latest code.

if yes, you should get back to us with details about the machine that you
are running on. people have been running calculations on system much larger
and been able to read and write restarts.

cheers,
    axel.

p.s.: have you checked, that you have not exhausted your disk quota or so?