Hi all,
Few simulations of mine will take even a day or more to finish.
When I am trying monitor the storage space on my hard-disk then it is being filled like 10GB/hr approximately but my post/directory is only 1GB .
Is LIGGGHTS writing some temporary files while I am running the code and kills those files after the simulation is DONE ?
This is a serious problem for me, as I am not able to launch multiple runs and if the run is for longer duration, CODE assumes as if there is no space to write DUMP files and gets killed.
Please help.
Thank you.
richti83 | Wed, 04/22/2015 - 10:08
standalone or coupled
are you doing a coupled CFDEM Simulation (in this case a lot of files are written to CFD folder and removed by Allrun after simulation is done) or a stand-a-lone LIGGGHTS simulation ?
Liggghts itself does not write any file when you do not explicitly use dump or fix print nor it delates any file after simulation finished.
Maybe you have a swap file on your root file system which gets blown up, check your real and virtual memory.
ravirasoon1991 | Wed, 04/22/2015 - 21:57
Mine are stand-a-lone
Mine are stand-a-lone LIGGGHTS simulations.
I found the problem.
Is there a way to overwrite the dump file at every time step?
I tried with "dump_modify dmp1 append no" which is explained in the doc/manual as "new dump file will be created which will overwrite an existing file with the same name" but I dint find this to be helpful for my case (file was not being overwritten)
richti83 | Thu, 04/23/2015 - 08:48
dump command
do you use dump filename*.extension or filename.extension ?
ravirasoon1991 | Thu, 04/23/2015 - 16:32
I use dump filename.extension
I use dump filename.extension.
I tried with filename*.extension and I was still not overwritting.
ravirasoon1991 | Thu, 04/23/2015 - 16:33
I use dump filename.extension
I use dump filename.extension.
I tried with filename*.extension and it was still not overwritting.
richti83 | Fri, 04/24/2015 - 12:48
whats going in there
a) filename.extension generates one big file with all timesteps included
b) filename*.extension generates a new file every dump-interval.
About hard disk storage:
every particle stores 19 values (id type type x y z ix iy iz vx vy vz fx fy fz omegax omegay omegaz radius) with the default dump command in the examples
every item has a length of 1 to 10 characters. One char is 1 byte, so a line for one particle with 85 characters (mean length in my example) has 85 bytes.
100 particle=8kb per dumpfile.
1e6 particle=8MB per dumpstep
Imagine you dump every 1000 iterations with a timestep of 1e-5 this is 0.001 or 100 files per second simtime. With 1e6 particles you need 800 MB per second.
You can now 1st reduce the items in the dump statement (you don't need ix/iy/iz, and mostly you are not interested in omaga and 2nd dump only every 0.01 second enlarging the dump every number.
I don't understand why your hard disk gets filled with 10GB/hr when the post folder doesn't grow but I think it's not related to liggghts itself. Can you append your inputdeck to the first post that I can have a look whats going on there ?
Cheers,
Christian.
ravirasoon1991 | Sat, 04/25/2015 - 01:44
Thanks for your reply
I am trying to write the contactarea (particle-particle and particle-wall) for each dump-interval.
When I use 1000 particles, I will have lots of data to be written since 10000 particles case is 10000*10000 operation. I will be a huge file by the end of by simulation(10GB/hr).
If I dump for every 1000 iterations instead of 1 then iggghts code will not calculate few things for me, that's the reason I have to dump for each iteration. And if I overwrite this dump file I might not have any problem.
So my question is, Is there a way to overwrite the dump-file for every iteration?
I am sorry I am not supposed to share my input file.
richti83 | Sat, 04/25/2015 - 17:55
quick & dirty hack
as a quick hack you could use
run $N pre no post no every 1 "shell rm post/dumpfile.liggghts"But I would try to get rid of the major problem. Do you use compute pair/gran/local area to compute the contactarea or do you have an own way of doing this ?
Did you try to use the thermo style to force re-evaluation of your variables ? Think about sending the console output to /dev/null than the terminal should be silend. also disable logging by echo none.
I can really better help you when you would share even then lines which are related to your special problem.
ravirasoon1991 | Mon, 04/27/2015 - 03:58
Hey thanks for the hack :D
Hey thanks for the hack :D
I am using compute pair/gran/local to calculate the contactArea and little modifications in the code, but not a professional way.
I think your hack fixed my problem.
ravirasoon1991 | Fri, 05/22/2015 - 21:57
Hey Christian,
Hey Christian,
"shell rm post/*" does'nt work but "shell gzip post/*" works .
Is there a way to do it?
Thanks