[GE users] Questions

[Reuti]

    [ The following text is in the "windows-1252" character set. ]
    [ Your display is set for the "ISO-8859-10" character set.  ]
    [ Some special characters may be displayed incorrectly. ]

Hi Dale,

the np_load_avg is not the number of slots used on a machine (neither by 
queue, nor in total), but the number of processes eligle for execution 
by the CPU. This is not bound to a queue, but the np_load_avg is the 5 
minutes average of the system load, which you will also see at the 
second load entry in "uptime". In addition, np_load_avg is the load_avg 
devided by the number of CPUs in the machine.

So your setup should accept unlimited jobs, until the value is above 
1.75 - then the queue is send to alarm state and won't accept any 
further jobs.

To avoid having too many jobs on a machine when they just started (until 
the load_avg reflects the activity of the just queued jobs), the 
scheduler knows a "job_load_adjustments", which will put some artificial 
load in the specified entries, to avoid this to happen (see "man 
sched_conf" for more).

Cheers - Reuti

Schmitz Dale M Contr 20 IS/INPG wrote:
> I?ve been trying to decipher the meaning of some basic information 
> attached to the queues on my grid engine.  I?m still dealing with the 
> basics, so this question is along those lines.
> 
>  
> 
> Load_threshold: All my queues have a default load_threshold of 1.75.  As 
> I understand it, if the engine sees more than 1.75 jobs assigned to that 
> queue, it won?t assign any more.  If I have that correct, no more than 1 
> job will be assigned to that queue, regardless of the fact that I have 
> two slots for that queue. 
> 
>  
> 
> Am I interpreting this correct?
> 
>  
> 
> TIA
> 
> Dale
> 


---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net