[GE users] Missing slots

Reuti reuti at staff.uni-marburg.de
Wed Aug 1 14:24:28 BST 2007


Hi,

Am 01.08.2007 um 15:00 schrieb Heywood, Todd:

> Never mind :-). It turns out these hosts were absent from  
> @publichosts due
> to some miscommunication here. Although all hosts should have 4  
> slots by
> default already "(qconf -se <host>" shows slots=4).

these are two different limits (queue and exechost), and the lower  
one will be enforced. The definition in the exechost makes of course  
mostly sense, if you have more than one queue on a machine, and the  
slot count across all these various queues would exceed the number of  
installed cores there.

-- Reuti


> Todd
>
>
> On 8/1/07 7:26 AM, "Reuti" <reuti at staff.uni-marburg.de> wrote:
>
>> Am 31.07.2007 um 18:18 schrieb Heywood, Todd:
>>
>>> I have a situation where SGE seems to only see one slot per host
>>> even though
>>> I have defined 4 slots per host. There is a single queue on these
>>> hosts.
>>> Here is an example, showing 1 slot available (0 used) on blade10:
>>>
>>> [root at bhmnode2 bh_admin]# qhost -h blade10 -q
>>> HOSTNAME                ARCH         NCPU  LOAD  MEMTOT  MEMUSE
>>> SWAPTO
>>> SWAPUS
>>> -------------------------------------------------------------------- 
>>> --
>>> ------
>>> ---
>>> global                  -               -     -       -
>>> -       -
>>> -
>>> blade10                 lx24-amd64      4  0.00    3.9G   54.2M
>>> 1.0G
>>> 20.8M
>>>    public.q             BIP   0/1
>>> [root at bhmnode2 bh_admin]#
>>>
>>>
>>> Now, in the definition of public.q, I have 4 slots per host
>>> (blade10 is in
>>> the publichosts hostgroup) defined in this line:
>>>
>>> slots                 1,[@publichosts=4]
>>
>> Just to be sure: can you please post the result of:
>>
>> qconf -shgrp @publichosts
>>
>> And if the workload allows it: you can just define 4 as default, and
>> check whether the output qhost -q changes. Is it the same as in qstat
>> -f regarding the slots?
>>
>> -- Reuti
>>
>>>
>>> And the local host config is also defined with 4 slots:
>>>
>>> [root at bhmnode2 lx24-amd64]# qconf -se blade10
>>> hostname              blade10
>>> load_scaling          NONE
>>> complex_values        virtual_free=3.8G,slots=4
>>> <<snip>>
>>>
>>> The strange thing is that this is the case for only about 5% of the
>>> hosts in
>>> the publichosts hostgroup.
>>>
>>> Any ideas? Thanks.
>>>
>>> Todd Heywood
>>>
>>>
>>> -------------------------------------------------------------------- 
>>> -
>>> To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
>>> For additional commands, e-mail: users-help at gridengine.sunsource.net
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
>> For additional commands, e-mail: users-help at gridengine.sunsource.net
>>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
> For additional commands, e-mail: users-help at gridengine.sunsource.net

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




More information about the gridengine-users mailing list