Platform LSF Administration Guide Version 6.2

Chapter 28
Running Parallel Jobs
Administering Platform LSF
443
Changing PROCLIMIT
If you change the PROCLIMIT parameter, the new processor limit does not affect
running jobs. Pending jobs with no processor requirements use the new default
PROCLIMIT value. If the pending job does not satisfy the new processor limits for the
queue, it remains in PEND state, and the pending reason changes to the following:
Job no longer satisfies queue PROCLIMIT configuration
If PROCLIMIT specification is incorrect (for example, too many parameters), a
reconfiguration error message is issued. Reconfiguration proceeds and the incorrect
PROCLIMIT is ignored.
MultiCluster
Jobs forwarded to a remote cluster are subject to the processor limits of the remote
queues. Any processor limits specified on the local cluster are not applied to the remote
job.
Automatic queue selection
When you submit a parallel job without specifying a queue name, LSF automatically
selects the most suitable queue from the queues listed in the DEFAULT_QUEUE
parameter in
lsb.params or the LSB_DEFAULTQUEUE environment variable.
Automatic queue selection takes into account any maximum and minimum
PROCLIMIT values for the queues available for automatic selection.
If you specify
-n min_proc,max_proc, but do not specify a queue, the first queue
that satisfies the processor requirements of the job is used. If no queue satisfies the
processor requirements, the job is rejected.
Example
For example, queues with the following PROCLIMIT values are defined in
lsb.queues:
◆ queueA with PROCLIMIT=1 1 1
◆ queueB with PROCLIMIT=2 2 2
◆ queueC with PROCLIMIT=4 4 4
◆ queueD with PROCLIMIT=8 8 8
◆ queueE with PROCLIMIT=16 16 16
In lsb.params: DEFAULT_QUEUE=queueA queueB queueC queueD queueE
For the following jobs:
% bsub -n 8 myjob
LSF automatically selects queueD to run myjob.
% bsub -n 5 myjob
Job myjob fails because no default queue has the correct number of processors.