[STALLED] PP-759: possibility to disable job-wide limit enforcement for host exclusive jobs

PP-759 - Getting issue details... STATUS

Overview:

  • If we use job-wide limit enforcement, the host exclusive jobs are also killed once the requested resources are exceeded even though the host is fully dedicated to the job. This is a mom's config variable for controlling this behavior.

Interface 1: enforce_on_exclhost

  • Visibility: Public
  • Change Control: stable
  • Synopsis: mom's config variable; variable can control whether the job-wide limits (ncpus /burst/, ncpus /sum/, mem, vmem ) are enforced on host exclusive jobs
  • Details: Mom's config variable. The type is boolean. Allowed values are yes/true/on/1 or no/false/off/0. This variable can be written into PBS_HOME/mom_priv/config file. Default value is true, which means all enforcement are applied to host exclusive jobs. Set to false means: Do not kill a host exclusive job if the ncpus (burst or sum), mem or vmem are exceeded.  -lplace=exclhost Other limits (like walltime or cput) are still enforced.