<style>@font-face{font-family:Calibri;panose-1:2 15 5 2 2 2 4 3 2 4;}</style><font face="Calibri"><p dir="ltr">IBM Knowledge Center says this:<br>
Job slot</p>
<p dir="ltr">A job slot is a bucket into which a single unit of work is assigned in the LSF system.</p>
<p dir="ltr">Hosts can be configured with multiple job slots and you can dispatch jobs from queues until all the job slots are filled. You can correlate job slots with the total number of CPUs in the cluster.</p>
<p dir="ltr">...Meaning you can slot one machine with a dozen slots and another machine of the same size with only one, the software should figure it out.I<br>
I hope this is helpful.<br>
Jonathan Engwall</p>
<br><br>On March 19, 2019, at 9:17 AM, Robert Taylor <rgt@wi.mit.edu> wrote:<br><br><br></font><div dir="ltr"><div dir="ltr" class="gmail_attr" style="color:rgb(80,0,80)">Hi All. We are on a old version of LSF where I am, lsf8.<br></div><div dir="ltr" style="color:rgb(80,0,80)"><div dir="ltr"><div dir="ltr"><div>We are trying to implement job packing on a particular queue, and I find some references online to usingĀ </div><div><br></div><div>RES_REQ = order[-slots]</div><div><br></div><div>in the lsb.queues file, andĀ </div><div>SCHED_PER_JOB_SORT=Y in lsb.params<br></div><div><br></div><div>However, this doesn't work, as they appear to only be supported on newer versions of lsf.</div><div>Can any lsf old-schoolers out there give me some recommendations on how to accomplish this?</div><div>They would like nodes to fill up with jobs before spreading them across more nodes, to leave more possible nodes free for high core count or exclusive jobs.</div><div><br></div><div>Thanks.</div><div><br></div><div>rgt</div></div></div></div></div>