HP XC System Software Administration Guide Version 4.0

This command searches through a list of nodes with the lsf service until it finds a node to run
LSF with SLURM.
Alternatively, you can invoke the following command to start LSF with SLURM on the current
node:
# controllsf start here
16.5.2 Shutting Down LSF with SLURM
At system shutdown, the /etc/init.d/lsf script ensures an orderly shutdown of LSF with
SLURM.
You can use the controllsf command, as shown here, to stop LSF with SLURM regardless of
where it is active in the HP XC system:
# controllsf stop
16.6 Controlling the LSF with SLURM Service
You can use the service command to start or stop the LSF with SLURM service on the HP XC
system, or to obtain the system's current status:
• service lsf start
This command is primarily of interest for automated startup. If the current node is the
primary LSF execution host, it sets the state to RUNNING, then starts LSF with SLURM unless
it is already running somewhere on the HP XC system. If the node is not the primary LSF
execution host, it ignores the command.
• service lsf stop
This command stops the LSF with SLURM environment if it is running on the current node.
Invoking this command on the LSF execution host or on the head node shuts down the LSF
with SLURM environment regardless where it is on the HP XC system, and sets the state to
SHUT DOWN to prevent any attempt to fail over the LSF with SLURM service to another
node.
• service lsf status
This command reports the current state (UP or DOWN) of LSF with SLURM.
This command has the same function as controllsf status.
16.7 Launching Jobs with LSF with SLURM
You may not submit LSF with SLURM jobs as superuser (root). You may find it convenient to
run jobs as the local lsfadmin user. An example would be a job to test a new queue configuration.
The LSF with SLURM daemons run on one node only: the LSF execution host. Therefore, they
can dispatch jobs only on that node. The JOB_STARTER script, described in “Job Starter Scripts”
(page 192), ensures that user jobs execute on their reserved nodes, and that these jobs do not
contend for the LSF execution host.
Consider an HP XC system in which node n120 is the LSF execution host, and nodes n1 through
n99 are compute nodes. The following series of examples shows jobs launched without the
JOB_STARTER script with varied results.
Example 16-3 illustrates the launching of a job in its most basic form.
16.6 Controlling the LSF with SLURM Service 197