Runtime Resource Usage Limits

Runtime Resource Usage Limits

Runtime limits are hard limits enforced by LSF.

Kicks in if users allocated more resources than allowed

Also covers users who dont bother inputting anything in to the bsub script.

Edit files in CFM:

/etc/cfm/templates/lsf

Show any runtime limits

bqueues -l normal

Set runtime limits

Edit files lsb.queues and/or lsb.applications, and enable memlimit in the lsf.conf file:

Update: PCM 3.2: /opt/lsf/conf/lsf.conf
File: /etc/cfm/templates/opt/lsf/conf/lsf.conf

# Local alterations
#LSB_JOB_MEMLIMIT=Y
LSB_MEMLIMIT_ENFORCE=Y
LSF_UNIT_FOR_LIMITS=MB

Update: PCM 3.2: /opt/lsf/conf/lsbatch/[cluster_name1]/configdir/lsb.queues
File: /etc/cfm/templates/lsf/conf/lsbatch/[cluster_name1]/configdir/lsb.queues



Begin Queue
QUEUE_NAME   = normal
PRIORITY     = 30
NICE         = 20
MEMLIMIT     = 1024 # 1MB in KB: default is KB
# Memlimit can take two args, one for default mem limit and one for max limit
MEMLIMIT     = default_limit max_limit
BACKFILL     = Y
...
DESCRIPTION  = For normal low priority jobs, running only if hosts are \
lightly loaded.
End Queue

Restart LSF

addhost -u
badmin hrestart all (then y)

Faster checks for jobs exceeding limits

Play with the SBD_SLEEP_TIME parameter (lsb.params)