[torquedev] r3806, auto_node_np

Ken Nielson knielson at adaptivecomputing.com
Fri Jul 2 12:17:03 MDT 2010


On 07/02/2010 12:10 PM, Garrick Staples wrote:
> Just a quick note regarding Ken's r3806 patch, here is his very good changelog note:
>
> When the server parameter auto_node_np is set to true it is suppose to set the
> number of processors of a node to the value returned by the MOM in the ncpus
> value as returned in pbsnodes. If the configured processor value is less than
> ncpus the value is adjusted but if it is greater the value was not adjusted.
> This fix enables pbs_server to adjust processor values down as well as up.
>
>
> However, the previous behaviour of only increasing np wasn't a bug. It was done
> on purpose because people wanted to oversubscribe nodes.
>
> I don't care either way, but someone out there might bring it up.
>
>    
>
>
> _______________________________________________
> torquedev mailing list
> torquedev at supercluster.org
> http://www.supercluster.org/mailman/listinfo/torquedev
>    
Garrick,

Thank you for the heads up.

If someone brings it up I will argue that if you want to oversubscribe 
nodes then why would you want to use auot_node_np. Its purpose is to 
automatically configure the number of processors to their real values.


Ken
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.supercluster.org/pipermail/torquedev/attachments/20100702/9a25f08c/attachment.html 


More information about the torquedev mailing list