Hi,

Am 10.01.2013 um 09:49 schrieb Arnau Bria:

> On Wed, 9 Jan 2013 20:29:12 +0100
> Reuti Reuti wrote:
> 
>> Hi,
> Hi Reuti,
> 
>>> the example shows 1000 units and halflife of 7 days:
>>> 
>>> 500 after 7 days
>>> 250 after 14 days
>>> 125 after 21 days
>>> 62.5 after 28 days
> 
>> You can only reset it in `qmon` completely.
> 
> how many time windows is it taking into account? 5?

I found this, maybe the factor can be changed:

libs/sched/sge_support.c:      *decay_rate = - log(0.5) / (halftime * 60);


>> There is a similar command:
>> 
>> /usr/sge/utilbin/lx24-amd64/sge_share_mon -?
> I must play with that, but default output is horrible... hard to
> understand and to manage.

Why do you want to manage it? The internal bookkeeping is like a steering wheel 
and the result maybe checked (i.e. scheduled jobs in the last couple of weeks) 
to proof whether the desired distribution was achieved in the accounting.

-- Reuti


> 
>>> I've configured (using qconf):
>>> usage_weight_list
>>> cpu=1.000000,mem=1.000000,io=0.000000
>>> 
>>> but when I open qmon->Policy->Share tree policy->botom-right arrow I
>>> see "Share tree policy parameters" (CPU,MEM,IO) all at 33%.
>> 
>> This happens if the sum of the values is larger than 1 in total, and
>> it's 2 in your case.
> 
> GReat! solved!
> 
>> -- Reuti
> Again, many thanks for your reply.
> 
> Cheers,
> Arnau


_______________________________________________
users mailing list
[email protected]
https://gridengine.org/mailman/listinfo/users

Reply via email to