Restart on a Cluster


I have some issues using the restart command with LAMMPS, especially on a Cluster.
I am preparing a system, going through some simulations steps. Between each step, I check several quantities, in order to apply modification to the next step. Because I want to keep the last state AND the velocities, I tried to use the restart command instead of write_data.

While doing a read restart, the run is incredibly slow. My guess is that the restart command is creating a file optimized for the current procs and nodes and if I’m using differents settings/procs/nodes for the next step, it won’t be able to run properly. Am I wrong ?

Is there anyway to overcome this or should I keep the same submission settings and try to get the same partition for every run ?

Thank you for your help,


When reporting an issue like this, you need to also report what LAMMPS version you are using and also provide a minimal (not your real) input deck (ideally done by modifying one of the example inputs) that can quickly reproduce the situation.

If that is not feasible, you should at least provide some log file(s) confirming your observations.