[torqueusers] Queue to node mapping with TORQUE only

[Davide Cesari]

Steve Young ha scritto:
> Hmm.... well according to the Torque Admin manual 
> (http://www.clusterresources.com/torquedocs21/4.1queueconfig.shtml#mapping) 
> it says:
> 
> "The simplest method is using default_resources.neednodes on an 
> execution queue, setting it to a particular node attribute. "
> 
> setting it to a node attribute I am assuming means a feature not a node 
> name.  I am using 2.2.1 but it was still the same before this when I was 
> using 2.0.0p7. I wonder if it does work with a node name if that should 
> be added to the documentation. Or did you add the node name as a feature 
> and it is really mapping a feature and not a node name. Just curious.
> 
> -Steve

Yes I read that page and I tried first by using a node attribute as you 
say (I mean the one listed in */torque/server_priv/nodes after the node 
name, just to be clear, "ext_net" in my case) and then, I am not really 
sure now, but looking at old server logs I think I received a message like

05/08/2008 14:21:50;0080;PBS_Server;Req;req_reject;Reject reply 
code=15062(Unknown node ), aux=0, type=RunJob, from Scheduler at maialinux
            ^^^^^^^^^^^^

when I tried to submit to that particular queue with -q, so I tried to 
use a node name instead of a node property and it magically worked. I 
must say, however, that my purpose was to setup a queue on a timeshared 
node, whose entry in nodes file looks like:

maialinux:ts ext_net

so maybe this is a special case!? Or maybe there is something wrong 
somewhere else in my setup?

	I hope the explanation is clear enough
		best regards, Davide	

> 
> On Jun 20, 2008, at 11:28 AM, Davide Cesari wrote:
> 
>> HI,
>>     I have Torque 2.1.10 (shipped with Fedora 8) without Maui/Moad and 
>> I noticed by trial-and-error that, in my case, 
>> "resources_default.neednodes" requires a node name, not a node 
>> property as specified in the documentation, maybe something has 
>> changed in more recent versions of Torque!?
>>     Davide
>>
>>
>> Steve Young ha scritto:
>>> I do it here with the resources_default.needsnodes and it works fine.
>>> in my server_priv nodes file I have
>>> host1 chem altix
>>> host2 physics altix
>>> then I could have three queue's....
>>> create queue chem
>>> set queue mac queue_type = Execution
>>> set queue mac Priority = 100
>>> set queue mac resources_default.neednodes = chem
>>> set queue mac resources_default.walltime = 24:00:00
>>> create queue physics
>>> set queue mac queue_type = Execution
>>> set queue mac Priority = 100
>>> set queue mac resources_default.neednodes = physics
>>> set queue mac resources_default.walltime = 24:00:00
>>> create queue altix
>>> set queue mac queue_type = Execution
>>> set queue mac Priority = 100
>>> set queue mac resources_default.neednodes = altix
>>> set queue mac resources_default.walltime = 24:00:00
>>> Doing that has worked fine here.  jobs in each queue go to the hosts 
>>> you would expect them to.
>>> Hope that helps,
>>> -Steve

-- 
============================= Davide Cesari ============================
Servizio IdroMeteorologico ARPA Emilia Romagna
Area Modellistica Numerica e Radarmeteorologia
  Phone/Fax: +39 051525926/+39 0516497501
  E-mail:    dcesari at arpa.emr.it
  Home page: http://www.webalice.it/o.drofa/davide/
  Address:   ARPA-SIM, Viale Silvani 6, 40122 Bologna, Italy
========================================================================