mpirun noticed that process rank 0 with PID 0 on node p2-gpu-10 exited on signal 9

[jie_yao2]

Dear VASP group,

I run VASP on gpus for long ab initio MD run for 20,000 ionic steps. (iron magnesium silicate liquid around 150 atoms)

The VASP report: Some of your processes may have been killed by the cgroup out-of-memory handler.

The memory is not enough, always after 1000 steps.

Is there a way to reduce memory or is it because the gpu hardware setup is not 100% correct ?

In my OUTCAR: total amount of memory used by VASP MPI-rank0 296452. kBytes.

thank you for your advice,

Jie

[henrique_miranda]

Could you share your input files so that we try to reproduce the issue?

The issue you are reporting sounds like a memory leak.
Do you have some monitoring tool that reports the memory usage during your calculation?
If the memory usage is increasing during the MD run is normally a sign of a memory leak.

See this related thread:
https://www.vasp.at/forum/viewtopic.php?f=3&t=18493