Hello all, I found a "bug" in slurmctld. The problem is simple.
If an administrator sets the MaxMemPerCPU in slurm.conf to 0 (maybe because thinking of disable it), and then a user who wants to set their memory requeriments like: srun --mem=1000 , a slurmctld crash takes place. The kernel message is: slurmctld[9152] trap divide error ip:43ea90 sp:7fa643bfa280 error:0 in slurmctld[400000+174000] It also generates a core dump but in this case I thing it's clear what is happening. Imho the slurmctld should control the divide by 0 case and act consequently, not crashing. Best regards.
