LSF_ENABLE_TMP_UNIT
Syntax
LSF_ENABLE_TMP_UNIT=Y | y | N | n
Description
The default value is LSF_ENABLE_TMP_UNIT=N|n. By default, the
LSF_UNIT_FOR_LIMITS parameter does not affect the
tmp resource, and the unit of the tmp
resource for
limits is MB.
When the LSF_ENABLE_TMP_UNIT=Y parameter is specified, the units that are
defined by the LSF_UNIT_FOR_LIMITS parameter also apply cluster-wide to the
tmp
resource. Applying limit units to the tmp resource
affects the following limits and resource requirements:- Resource requirements at the job-level (bsub), queue-level (lsb.queues), and application-level (lsb.applications).
- The TMP resource in the Limit section and RES_SELECT in the GuaranteedResourcePool section in the lsb.resources file.
- The load_index threshold in the lsb.queues and lsb.hosts file.
Example
When the LSF_ENABLE_TMP_UNIT=Y is set, and the unit for limits is specified as GB (LSF_UNIT_FOR_LIMIT=GB), the following job submission means mem=1 GB, swp=2 GB, tmp=3 GB, and tmp>3 GB":
bsub -R "rusage[mem=1,swp=2,tmp=3] select[tmp>3]"
Important:
Before you enable limits for the tmp resource, completely drain the cluster of all workload so that no running, pending, or finished jobs are in the system.
In the LSF multicluster capability environment, configure the same unit for all clusters.
Default
N | n