Description
Consider system signal for gentle run termination
Context: Wall-time for sims on HPC
Cluster usage is typically temporized, with simulations exceeding the wall-time being terminated
Typically, abruptly terminating a simulation results in data corruption, meaning the entire simulation output must be discarded.
Solution proposed:
“We have implemented a workaround based on system signals that allows the simulation to be stopped properly before the
wall time expires, allowing more flexibility and preventing data corruption
We believe this feature would be highly desirable for Geant4, as it would allow any simulation to be properly terminated upon
user request
Furthermore, this feature could be extended to provide additional information
We can share our implementation of this signaling with Geant4, which is in use in combination with Apptainer, in the Virgo cluster at GSI
Contacts: Enrico Pierobon, Christoph Schuy @GSI
See p. 10 of Giovanni's talk: https://indico.cern.ch/event/1554191/contributions/6712776/attachments/3157727/5609485/Geant4%202025%20GSantin%20Requirements%20space%20science%20engineering%20v1.01.pdf