🎉Community Raffle - Win $25

An exclusive raffle opportunity for active members like you! Complete your profile, answer questions and get your first accepted badge to enter the raffle.
Join and Win

What command do you use to get the number of GPUs available on a particular queue?

Suppose you have some nodes "n1", "n2", and "n3" associated with a particular queue named "gpu".  Each node has 4 GPUs.  What command do I use to query the queue to find out how many total GPUs are available on that queue, which, in this example, will be 12 GPUs, since there are 3 nodes associated with the queue and each node has 4 GPUs?

For example, I am setting the number of GPUs available on each node using the following commands.

qmgr -c 'set node n1 resources_available.ngpus=4'
qmgr -c 'set node n2 resources_available.ngpus=4'
qmgr -c 'set node n3 resources_available.ngpus=4'

 

I am associating nodes "n1", "n2", and "n3" with queue "gpu" using the following commands. 

</code><code class="java plain">qmgr -c&nbsp;<code class="java string">'create resource Qlist type=string_array, flag=h'<br /><br />grep Qlist /var/spool/pbs/sched_priv/sched_config<br />resources:&nbsp;"ngpus, ncpus, mem, arch, host, vnode, aoe, eoe, Qlist"

<code class="java string">qmgr -c&nbsp;'set queue gpu default_chunk.Qlist = gpuQ'

qmgr -c&nbsp;"set node&nbsp;n1 resources_available.Qlist='gpuQ'"<br /><code class="java plain">qmgr -c&nbsp;"set node n2 resources_available.Qlist='gpuQ'"
<code class="java plain">qmgr -c&nbsp;"set node n3 resources_available.Qlist='gpuQ'"

<br />What command do I use to report that queue "gpu" has 12 GPUs that are available, and also, if some of those GPUs become in use, what command do I use to show that "N" out of 12 GPUs are in use?

Find more posts tagged with

No comments on this post.