[slurm-users] detectCores() mess

[Chris Samuel]

On Tuesday, 12 December 2017 7:42:51 AM AEDT Mike Cammilleri wrote:

> I guess that when this happens, the load average in 'top' can show an
> extremely elevated number due to the fact that lots of processes are
> waiting to run - but in fact the overall availability of the node is still
> quite open as there are plenty of available cores left for other jobs.
> Would this be an accurate interpretation of the scheduling and load I'm
> observing? Are there impacts to the performance of the node when it is in
> this state?

I think that's a pretty accurate summary.  You just end up with a heap of 
processes fighting over the same core(s) they are constrained to and polluting 
each others cache.

I would have thought that should have minimal impact on other processes on the 
node constrained to different cores.

Good luck!
Chris
-- 
 Chris Samuel  :  http://www.csamuel.org/  :  Melbourne, VIC