Hello,
I am trying to model a core with TRISO fuel explicitly modeled, with the openmc.model.create_triso_lattice method, and with a small pitch, the memory usage goes way up and the calculation is aborted by the server, which I understand as the number of cells becomes too large. But with a larger pitch, it is getting stuck in the initialization phase, apparently when writing the summary.h5 file, which I don’t understand.
Preparing distributed cell instances...
Reading plot XML file...
Writing summary.h5 file...
=>> PBS: job killed: walltime 86495 exceeded limit 86400
On a smaller scale (for instance, one fuel assembly) there is no problem. Any advice is much appreciated, thanks!
I was able to get past this problem by disabling the generation of the summary.h5 output. It was getting fairly large (> 5Gb) and still being written after 24 hours. Not sure if it is expected for complicated geometries?
Glad you were able to figure that out as that was what I would have suggested anyway. It’s not entirely unexpected if you have a very complicated geometry that it could take some time to write that file, although >24 hours seems pretty excessive. I’ll try playing around with some big models I have on hand to see if I can identify anything obvious that could speed it up.
For anyone coming across this, disabling generation of the summary.h5 file can be done with:
settings = openmc.Settings()
settings.output = {'summary': False}