MPI error and particles lost rapidly

[SunnyQbang]

Hi every one!
I create a new environment using conda to compile dealii.9.6 to compile aspect in the supercomputer .And i use the module mpi gun 4.1.4. But there is a problem when get in the step30.

And i use the free slip at the other 7 except top,top is freesurface.I set a left boundary velocity like that ,so if it is the reason why the particle decrease rapidly in a few step?

I use 200million particles and at step30 it lost about 210hundred,

i dont konw how to explain it.And it indeedly dont occur the error i mentioned before

Error in particle property_pool

Dear all, Recently,i want to use particles to trace my compositional elements. I ran this program on two identical servers, both using 40,000,000 particles. The only difference is in the parameter set Cells along circumference, where one used the default value of 12 and the other used 4. The error occurs in the program that used the value of 4. What could be the reason for this error, and how can it be resolved? [16dce9844ccfc3323f76d303ab3d491]

I would appreciate any reply!

[bangerth]

@SunnyQbang Let’s do one question at a time. Can you post the complete input file that caused the error, and show us the first 20 or so lines of the ASPECT output?
Best
W.

[bangerth]

@SunnyQbang This is quite a large number of MPI processes (more than 3000). Do you actually need this many? At least in the initial time steps your log file shows, you do not have very many degrees of freedom, and could use many fewer MPI processes. The manual has a section on how many processes you generally need.

I mention this because it is much easier to debug problems with smaller setups. Can you see if you can reproduce the MPI failure with a small number of processes? The same applies to the particle part of your question.

Best
W.