IBM Power Systems 775 Manual page 298

For aix and linux hpc solution
Table of Contents

Advertisement

To monitor the status of the jobs, issue the llstatus command to one of the service nodes,
as shown in Example 5-19.
Example 5-19 Monitoring the status of the jobs
xdsh f01sv01-v llstatus
Important: Because the draining process allows the jobs to complete, this step continues
as long as it takes for the longest running job to complete. Familiarity with the jobs and the
amount of time the job run helps determine the length of this task.
Method 2 - Stopping LoadLeveler jobs
Example 5-20 shows how to keep any new jobs from starting in LoadLeveler.
Example 5-20 Draining LoadLeveler jobs on the compute nodes and service nodes
for n in `nodels compute`
do
llctl -h $n drain startd
done
xdsh service -v llrctl drain
Wait for running jobs to complete, or alternately, if the jobs are terminated and restarted, flush
the jobs on the compute nodes by entering the commands that are shown in Example 5-21.
Example 5-21 Flushing the jobs in the compute node
xdsh compute -v llrctl flush
Important: In the job command file, the restart=yes flag must be specified. If the flag is
not specified, it is similar to llcancel. The jobs that are running on a node are gone
permanently after you flush the node.
To monitor the status of the jobs, issue the llstatus command to one of the service nodes as
shown in Example 5-22.
Example 5-22 Monitoring the status of the jobs in a service node
xdsh f01sv01-v llstatus
Stopping LoadLeveler
Shutting down LoadLeveler early in the process reduces any chances of job submission, and
eliminates any LoadLeveler dependencies on the cluster.
As described in "Cluster shutdown assumptions" on page 282, it is necessary to drain or
cancel all jobs and remove users from the system. It is assumed that all jobs are drained or
canceled by using the steps that were described previously in this section. Because there are
no jobs active in the system, this step describes shutting down LoadLeveler down.
LoadLeveler must be stopped on all compute nodes and service nodes, as shown in
Example 5-23 on page 285.
284
IBM Power Systems 775 for AIX and Linux HPC Solution

Advertisement

Table of Contents
loading

Table of Contents