segmentation fault

[aureliayang]

Dear all,

I frequently encountered such errors on the supercomputer. I am not sure if this is due to parflow code or the environment.
Another frequently encountered issue is that all log files (kinsol etc.) are not updated anymore but the job is still running (for several hours).
Could anyone give some thoughts?

Thank you very much!

Chen

[smithsg84]

Hi,

Sorry you are having problems. The error looks like it might be related to MPI based on the stack trace provided. The hanging is also a sign that the MPI communication is not working correctly. Building ParFlow with debugging enabled might provide more information on where the error is occuring. In your cmake for building you can turn on debugging with

cmake -DCMAKE_BUILD_TYPE=Debug

This should add file and line number information to the segmentation fault output which may aid in tracking source of the problem.

[aureliayang]

Thank you, Steve! I build parflow again and submitted the job again.
I will keep you posted once the job crashed again

[aureliayang]

We found the reason for this segmentation fault and it is probably environment/version dependent. There is a bug in lines 373-409 which is introduced due to #588.

It can be seen that clm(t)%topo_mask(1) will be NaN for inactive cells and thus porosity(l) would not be accessible.

My student will submit the pull request soon. We had this issue randomly on a large supercomputer and it is totally blocked on the small cluster in the group.