ICC Home Privacy and Legal Notice LC User Documents Banner

UCRL-WEB-201386

SLURM Reference Manual


Resource Control

-I (uppercase, --immediate)
exits if requested resources are not available at once (by default, SRUN blocks until requested resources become available).
--network=type
(AIX systems only, where SLURM has replaced LoadLeveler) specifies the communication protocol to use (normally set with environment variable SLURM_NETWORK; see that section below). For systems with an IBM Federation switch (Purple, UM, UV, UP), the choices for type include IP, SN_ALL, SN_SINGLE, BULK_XFER, and the adapter names used as values for POE environment variable MP_EUIDEVICE (see the "Task Communications" section in the POE User Guide).
-O (uppercase oh, --overcommit)
overcommits CPUs. By default, SRUN never allocates more than one process per CPU. If you intend to assign multiple processes per CPU, you must invoke the -O option along with -n and -N (thus -n16-N4-O together allow 2processes/CPU on the 4 allocated 2-CPU nodes). Even with -O, SRUN never allows more than MAX_TASKS_PER_NODE tasks to run on any single node.
--propagate[=rlimits]
specifies which of the modifiable (soft) resource limits rlimits to propagate to the compute nodes and hence apply to this job (without rlimits, propagates all resource limits).
-W sec (uppercase, --wait=sec)
waits sec seconds after the first task terminates before terminating all remaining tasks (default for sec is unlimited). Use -W to force an entire job to end fairly quickly if any one task terminates prematurely.



Navigation Links: [ Document List ] [ HPC Home ] [ Next ]