[torqueusers] node's np still not right

Ken Nielson knielson at adaptivecomputing.com
Thu Mar 10 08:46:49 MST 2011


On 03/10/2011 08:33 AM, skip at pobox.com wrote:
> You may recall that I had trouble with Torque (2.4.8) overriding the number
> of processors I had explicitly set in qmgr.  One of my desktop machines has
> four cores.  I'm happy to devote at most two of them to the cause, but the
> auto_node_np default setting of True seemed to thwart my efforts at
> controlling this value.  I set auto_node_np tp False a couple days ago, set
> np = 2 and things seemed to work well at first.  Overnight, the
> pbs_server-hosting machine was rebooted.  This morning I noticed, that even
> though auto_node_np is still set to False:
>
>      Qmgr: p s
>      ...
>      set server auto_node_np = False
>      ...
>
> The number of processors on my desktop machine was once again reported to be
> four:
>
>      Qmgr: print node udesktop264.wacker
>      ...
>      set node udesktop264.wacker np = 4
>      ...
>
> I set it back to 2, but think there is either some other setting I am
> failing to take care of, or there is a bug in Torque which allows it to
> continue to ignore my commands, at least across pbs_server restarts.
>
When pbs_server rebooted it read the nodes file again. If np was not 
changed in the node file then the node would have np=4 on the reboot. Is 
there a reason you do not want to directly change the nodes file?

Ken


More information about the torqueusers mailing list