From 7e1811130b616e4b52b9d3478f87766a68ef1c78 Mon Sep 17 00:00:00 2001
From: Morris Jette <jette@schedmd.com>
Date: Thu, 15 Sep 2011 17:23:37 -0700
Subject: [PATCH] Describe mechanism to reserve CPUs rather than whole nodes
Update reservation web page to describe mechanism to reserve CPUs rather than whole nodes and provide an example.
---
doc/html/reservations.shtml | 28 ++++++++++++++++++++++++++--
1 file changed, 26 insertions(+), 2 deletions(-)
diff --git a/doc/html/reservations.shtml b/doc/html/reservations.shtml
index 2c80c74e7bd..f94179df5a9 100644
--- a/doc/html/reservations.shtml
+++ b/doc/html/reservations.shtml
@@ -106,7 +106,7 @@ ReservationName=root_5 StartTime=2009-02-04T16:22:57
Users=root Accounts=(null)
</pre>
-<p>Our final example is to reserve ten nodes in the default
+<p>Our next example is to reserve ten nodes in the default
SLURM partition starting at noon and with a duration of 60
minutes occurring daily. The reservation will be available
only to users alan and brenda.</p>
@@ -131,6 +131,28 @@ reservation. Note that the reservation creation request can also
identify the partition from which to select the nodes or _one_
feature that every selected node must contain.</p>
+<p>On a smaller system, one might want to reserve specific CPUs rather than
+whole nodes. While the resolution of SLURM's resource reservation is that of
+whole nodes, one might configure each CPU as a license to SLURM and reserve
+those instead (we understand this is a kludge, but it does provide a way to
+work around this shortcoming in SLURM's code). Proper enforcement then requires
+that each job request one "cpu" license for each CPU to be allocated, which
+can be accomplished by an appropriate job_submit plugin. In the example below,
+we configure the system with one license named "cpu" for each CPU in the
+system, 64 in this example, then create a reservation for 32 CPUs. The
+user developed job_submit plugin would then explicitly set the job's
+licenses field to require one "cpu" for each physical CPU required to satisfy
+the request.</p>
+<pre>
+$ scontrol show configuration | grep Licenses
+Licenses = cpu*64
+
+$ scontrol create reservation starttime=2009-04-06T16:00:00 \
+ duration=120 user=bob flags=maint,ignore_jobs \
+ licenses=cpu*32
+Reservation created: bob_5
+</pre>
+
<h2>Reservation Use</h2>
<p>The reservation create response includes the reservation's name.
@@ -252,9 +274,11 @@ the highest level rather than the actual level of time-slicing when
considering the initiation of jobs.
This will prevent the initiation of some jobs which would complete execution
before a reservation given fewer jobs to time-slice with.</li>
+<li>Add support to reserve specific CPU counts rather than require whole
+nodes be reserved (work around described above).</li>
</ol>
-<p style="text-align: center;">Last modified 16 March 2011</p>
+<p style="text-align: center;">Last modified 15 September 2011</p>
<!--#include virtual="footer.txt"-->
--
GitLab