To whomever is considering answering,
I am currently running a modified icoFoam (added tracer transport) application in parallel on two of my computer's processors. I am coming across crash issues sometimes, where mpirun suddenly crashes with no error report whatsoever. This doesn't always happen, but what seems like out of nowhere, the application sometimes fails. It just stops at an random time step.
I would expect that if it is a memory error, an error report would appear as it is a standard issue with applications.
If you have come across this issue, please let me know what you did in circumventing this annoying issue.
Thanks in advance.
Hello Christopher,
you are right, it is unusual that OpenFOAM doens't print an error report. Maybe you cant try to run your simulation with the "> log" command and see if icoFoam generate an error message that way.
Dear Robert,
Thank you for your response. In fact, I've tried to run the application by pushing it as a background operation (with the operation reports stored in a .txt file) and on terminal, I never receive an error report. This issue is quite random too, it happens unexpectedly and doesn't have to do with numerical stability.
For those who are interested, I believe the problem was largely due to the temperature buildup on my CPUs. Thus, this issue was most probably due to an automated CPU break and not an OpenFOAM based break. This can explain why no error log was projected.
It might not be considered as typical crash, but no error output with mpirun can also occur, when deadlock happens. Thus, one CPU is waiting for the other. The application can be seen, when top is run and using mpirunDebug can show SIGHUP signal.
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running