Hello everyone,
Recently I have been working on a simulation to extract the contact angle between amorphous silica and a droplet of water, but my simulation has been running for 5 days.
But I have the opportunity to simulate on a supercomputer, but I have no idea how to do this, if someone knows a blog, video, workshop or something that helps me with this issue please let me know.
Thanks
M. Lerdon
I am not sure that it is a LAMMPS specific question, however, you need to be more specific in order to obtain some answers.
For instance, you should say what supercomputer type is it? Is it Cray or are those regular nodes that are connected through Infiniband network, etc.
My perception though it would be much more desirable for you to contact the administrator who possibly already have some experience with questions of this kind
Best wishes,
Lukasz Baran
Ph.D. Candidate
Department of Theoretical Chemistry
Institute of Chemical Sciences
Faculty of Chemistry
Maria Curie Skłodowska University in Lublin
ORCID: 0000-0003-1777-1998
ResearcherID: AAP-1152-2021
The easiest way to understand the procedure for performing simulations on a supercomputer is to contact the technical support that manages the use of the supercomputer via email. If it is a multi-user supercomputer of a research center or institution (e.g., university), there will probably be a website with a manual of some basic procedures to be carried out to execute a script (job).
If you are using a personal-use supercomputer in a professor’s group, there will likely be a graduate student who has more experience on the subject and can help you. If you are purchasing a supercomputer for the first time, I think your university’s IT department can help you build lammps by following the instructions in the software manual (https://docs.lammps.org/Build_basics.html).
The procedure for running a simulation of LAMMPS on a supercomputer is similar to that on a desktop: you will have an executable and you need to call it on a command line to run your script.
To use supercomputers of research centers or institutions, you usually need to write a project that a technical committee reviews, and may or may not approve your use of computing resources. Project submission information is usually also available on the website of the research center or institution.
In addition, you also need to know how files are uploaded and downloaded to the supercomputer (I explain this below).
The big point is that for you to run your LAMMPS simulation in supercomputers, you will have to submit your LAMMPS script in another script (called a job) which is tied to a queue management system. I believe that every supercomputer should have a queue management system. If the amount of processors requested in the job is available, your LAMMPS script will start to run. Note that the execution of your LAMMPS script may not be instantaneous, as it will depend on the position in the queue system. Examples of queuing systems are slurm, pbs… (you can search for them on the internet).
Assuming you have an accepted project (or know a university professor who has already provided you with resources), when contacting the technical support that manages a supercomputer, some of the basic questions you should ask are:
- Is there a user manual for the supercomputer? Ps: if it exists, you will probably find the answer to all the questions below.
- What queuing system is used in the supercomputer? Ps: Knowing what the system is (e.g., slurm, pbs), it is easier for you to write a job to run your LAMMPS simulation. You can find examples on the internet to use the indicated queuing system.
- What version of LAMMPS is installed on the supercomputer? Ps: you can usually ask the supercomputer administrators to install specific versions of LAMMPS in case the one you need is not available.
- Is there any standard script for using LAMMPS on the supercomputer? Ps: you are probably not the first user to use LAMMPS on the supercomputer, so there may be standardized software scripts.
- What command should you run (e.g. in Linux terminal) to connect to the supercomputer? Ps: when you use a cluster, you will run LAMMPS remotely and need a command to access the machine remotely (e.g., ssh command).
- What commands should you use to upload and download files on the supercomputer? Ps: usually an scp command is used.
Generally a job has the following information:
Hello,
A person who is working with me, help me with the commands, scripts, inputs, lines and all the necessary stuff to run the simulation.
I followed all the steps that he made to run his simulation, including upload the lammps carpet on the HPC with the “lmp_mpi”.
So in terms of procedure and tasks to do for the run, i think it"s ok.
So I started with the simulations and an error come up with this lines.
"…
./lmp_mpi: /lib64/libm.so.6: version GLIBC_2.29' not found (required by ./lmp_mpi) ./lmp_mpi: /home/modules/software/GCCcore/8.2.0/lib64/libstdc++.so.6:version GLIBCXX_3.4.26’ not found (required by ./lmp_mpi)
… "
I don"t know if the problem is that the supercomputer doesn"t have the library required and I need to ask the support team
to upload it or I compiled the lmp_mpi (executable for simulate in parallel) in a wrong way.
Please help me with this issue.
Regards
Here you can find more info about de supercomputer
https://wiki.fysik.dtu.dk/niflheim/Niflheim7_Getting_started
Matias Lerdon.
This is definitely something where you need to contact the user support of the center that you are running on.
Likely you didn’t follow the instructions they provide. This is a generic error (compiler and C library version used to link your LAMMPS executable are inconsistent with what is available on the machine that you are running on) and has nothing to do with LAMMPS but everything with the local machine setup and how your compiled/installed/used LAMMPS. Only a person local that that center can help you for real.
Thanks for the tip!
El El mié, 26 de ene. de 2022 a la(s) 19:56, Axel Kohlmeyer <[email protected]> escribió: