Re: [Condor-users] Condor & SMP configuration

[Steven Timm <timm@xxxxxxxx>]

Do you have NUMBER_OF_CLAIMED_CPUS added to STARTD_ATTRS
You have to do that otherwise it won't work.
Also see Matt's post that came later.. Anyone I know who has made
it work has done something like what is on that web page.

Steve Timm


On Wed, 1 Oct 2008, Andrea Borsic wrote:

> Dear All,
>
> I have been reading the user forums and trying to follow the proposed
> solutions, but I am still having troubles with the configuration of
> Condor for use on a SMP cluster. Each node on the cluster has 8 CPUs,
> and I am using the trick of presenting several types of slots in order
> to allow running jobs on 1,2,3,4,8 threads. The configuration file
> (attached below) has a logic that should prevent too many jobs to run on
> the single node.
>
> If I submit a mix of 8,4,2, and 1 threaded jobs, I get for example a 8
> threaded job running concurrently with three or four 2 threaded jobs on
> the same node, exceeding the number of CPUs available. The SLOTx_START
> expressions should prevent this, they seem to be ineffective though.
>
> Any suggestion is kindly appreciated.
>
> Regards,
>
> Andrea Borsic
>
>
>
> # here we compute the total number of cpus in use
> NUMBER_OF_CLAIMED_CPUS = \
> (\
> (1*(slot1_State =?= "Claimed")) + \
> (1*(slot2_State =?= "Claimed")) + \
> (1*(slot4_State =?= "Claimed")) + \
> (1*(slot5_State =?= "Claimed")) + \
> (1*(slot6_State =?= "Claimed")) + \
> (1*(slot7_State =?= "Claimed")) + \
> (1*(slot8_State =?= "Claimed")) + \
> (2*(slot9_State =?= "Claimed")) + \
> (2*(slot10_State =?= "Claimed")) + \
> (2*(slot11_State =?= "Claimed")) + \
> (2*(slot12_State =?= "Claimed")) + \
> (4*(slot13_State =?= "Claimed")) + \
> (4*(slot14_State =?= "Claimed")) + \
> (8*(slot15_State =?= "Claimed"))   \
> )
>
> # start single threaded jobs only if at least 1 cpu is available
> SLOT1_START = ($(NUMBER_OF_CLAIMED_CPUS) <= 7)
> SLOT2_START = ($(NUMBER_OF_CLAIMED_CPUS) <= 7)
> SLOT3_START = ($(NUMBER_OF_CLAIMED_CPUS) <= 7)
> SLOT4_START = ($(NUMBER_OF_CLAIMED_CPUS) <= 7)
> SLOT5_START = ($(NUMBER_OF_CLAIMED_CPUS) <= 7)
> SLOT6_START = ($(NUMBER_OF_CLAIMED_CPUS) <= 7)
> SLOT7_START = ($(NUMBER_OF_CLAIMED_CPUS) <= 7)
> SLOT8_START = ($(NUMBER_OF_CLAIMED_CPUS) <= 7)
> # start 2 threaded jobs only if at least 2 cpus are available
> SLOT9_START = ($(NUMBER_OF_CLAIMED_CPUS) <= 6)
> SLOT10_START = ($(NUMBER_OF_CLAIMED_CPUS) <= 6)
> SLOT11_START = ($(NUMBER_OF_CLAIMED_CPUS) <= 6)
> SLOT12_START = ($(NUMBER_OF_CLAIMED_CPUS) <= 6)
> # start 4 threaded jobs only if at least 4 cpus are available
> SLOT13_START = ($(NUMBER_OF_CLAIMED_CPUS) <= 4)
> SLOT14_START = ($(NUMBER_OF_CLAIMED_CPUS) <= 4)
> # start 8 threaded jobs only if all cpus are available
> SLOT15_START = ($(NUMBER_OF_CLAIMED_CPUS) == 0)
>
> MEMORY = 32768
> NUM_CPUS = 32
>
> SLOT_TYPE_1 = cpus=1, ram=1024
> SLOT_TYPE_2 = cpus=2, ram=2048
> SLOT_TYPE_3 = cpus=4, ram=4096
> SLOT_TYPE_4 = cpus=8, ram=8192
> NUM_SLOTS_TYPE_1 = 8
> NUM_SLOTS_TYPE_2 = 4
> NUM_SLOTS_TYPE_3 = 2
> NUM_SLOTS_TYPE_4 = 1
>
>
>
> _______________________________________________
> Condor-users mailing list
> To unsubscribe, send a message to condor-users-request@xxxxxxxxxxx with a
> subject: Unsubscribe
> You can also unsubscribe by visiting
> https://lists.cs.wisc.edu/mailman/listinfo/condor-users
>
> The archives can be found at:
> https://lists.cs.wisc.edu/archive/condor-users/

--
------------------------------------------------------------------
Steven C. Timm, Ph.D  (630) 840-8525
timm@xxxxxxxx  http://home.fnal.gov/~timm/
Fermilab Computing Division, Scientific Computing Facilities,
Grid Facilities Department, FermiGrid Services Group, Assistant Group Leader.