No output in log.lammps file

Dear lammps users
I ran lammps in school’s cluster. The simulation can run without error, but there is no output in log.lammps file. I have used the following command
thermo 100
So, there should be outputs in log.lammps file. My fist guess is that the log.lammps is too big and the cluster has limitation which does not allow to output the log file. Then I used the the same input with just 200 timesteps. Then there are output in the file.

But I used the same input which does not generate the log.lammps on cluster to run the simualtion on my workstation, I foud the trajectory file is 189078 kb which is much larger than the log.lammps with the size of 2988 kb. It seems that the problem is not related to size of log.lammps.

My questions are
(1)Is that the big size of log.lammps results in no output?If so, is there a command to split the log.file to make it smaller?
(2)If not, what results in the no output?

Sincerely
Fan Li

No, it is not because of the size of the log file but has to do with flush operation (writting back the contents from cache to main memory). You can use “thermo_modify flush yes” to invoke/enforce flush operation whenever the log file is written/updated on the cache. See thermo_modify (http://lammps.sandia.gov/doc/thermo_modify.html) command for more details.

Ray

Hi Ray
There is still no output in the log.lammps after I used the “thermo_modify flush yes” command. The attachment is my input. I found there is no output when the simulation is running, so I kill the simulation. But I think it is nothing to do with my problem,because “thermo_modify flush yes” command insures the output in that file is current, even if LAMMPS halts before the simulation completes. Do you know other reasons which may result in my problem?

in.coro-oh (2.51 KB)

Hi Ray
There is still no output in the log.lammps after I used the “thermo_modify
flush yes” command. The attachment is my input. I found there is no output
when the simulation is running, so I kill the simulation. But I think it is
nothing to do with my problem,because “thermo_modify flush yes” command
insures the output in that file is current, even if LAMMPS halts before the
simulation completes. Do you know other reasons which may result in my
problem?

if you are running on a cluster with a networked file system (NFS) for
home directories, then it can sometimes take some time until data
appears due to multiple levels of caching involved. using the flush
feature only flushes data to the NFS client. however, if you are
located on a node different from the one doing the i/o, such updates
may not show up for minutes. buffering is done in blocks of 4k or 8k
bytes and that corresponds to quite some time (a typical 80x25
character size xterminal holds 2k), thus it may need 2-4 screens full
of accumulated output before a buffer is committed. if a file is
closed regularly, then a forced synchronization happens (but again, it
will not automatically show up immediately). if a run is terminated
beforehand, the buffered data is usually lost. batch systems handle
output to the screen and particularly to the stderr channel
differently (by spooling to a local directory and the moving the
spooled files after the job is terminated), so those, especially
stderr, are usually (more) complete.

screen output is handled with more priority, if you run interactively.
most batch systems also support requesting an interactive session.

axel.