Hi,

Am 20.03.2014 um 18:15 schrieb Karun K:

> Our scheduler is configured to "least_used_host" policy depending on load 
> average and for PE environments its $pe_slots 
> Regular jobs are being allocated as expected but PE jobs are being filled up 
> before it moves to next available node.
> How can I configure PE jobs also to be round-robin? i.e all requested slots 
> in PE jobs have to be in the same host but jobs should be distributed rather 
> than filling up host. 

Is this issue limited to parallel jobs? For $pe_slots it should work, and I see 
you already defined a "job_load_adjustments".

-- Reuti


> Included our ge configs below , version 2011.11p1
> 
> Thanks,
> Karun
> 
> job-ID  prior   name       user         state submit/start at     queue       
>                    slots ja-task-ID 
> -----------------------------------------------------------------------------------------------------------------
>  124688 0.51929 STDIN      kk   r     03/13/2014 23:07:57 
> [email protected]            2        
>  124689 0.51929 STDIN      kk   r     03/13/2014 23:07:57 
> [email protected]            2        
>  124690 0.51929 STDIN      kk   r     03/13/2014 23:07:57 
> [email protected]            2        
>  124691 0.51929 STDIN      kk   r     03/13/2014 23:08:02 
> [email protected]           2        
>  124692 0.51929 STDIN      kk   r     03/13/2014 23:08:02 
> [email protected]           2           
>  124694 0.50500 STDIN      kk   r     03/13/2014 23:08:27 
> [email protected]           1        
>  124695 0.50500 STDIN      kk   r     03/13/2014 23:08:27 
> [email protected]            1        
>  124696 0.50500 STDIN      kk   r     03/13/2014 23:08:27 
> [email protected]           1        
>  124697 0.50500 STDIN      kk   r     03/13/2014 23:08:27 
> [email protected]            1   
> 
> [root@cluster ~]# qconf -ssconf
> algorithm                         default
> schedule_interval                 0:0:05
> maxujobs                          0
> queue_sort_method                 load
> job_load_adjustments              np_load_avg=3.0
> load_adjustment_decay_time        0:7:30
> load_formula                      np_load_avg
> schedd_job_info                   true
> flush_submit_sec                  0
> flush_finish_sec                  0
> params                            none
> reprioritize_interval             0:0:0
> halftime                          168
> usage_weight_list                 cpu=1.000000,mem=0.000000,io=0.000000
> 
> ----
> 
> [root@cluster ~]# qconf -sp threaded
> pe_name            threaded
> slots              9999
> user_lists         NONE
> xuser_lists        NONE
> start_proc_args    /bin/true
> stop_proc_args     /bin/true
> allocation_rule    $pe_slots
> control_slaves     FALSE
> job_is_first_task  TRUE
> urgency_slots      min
> accounting_summary FALSE
> 
> All nodes have identical complex configuration
> 
> [root@cluster ~]# qconf -se compute-4-3
> hostname              compute-4-3.local
> load_scaling          NONE
> complex_values        slots=30,h_vmem=120G
> load_values           arch=linux-x64,num_proc=30,mem_total=123136.023438M, \
>           -------------------truncated-----------------------
> processors            30
> user_lists            NONE
> xuser_lists           NONE
> projects              NONE
> xprojects             NONE
> usage_scaling         NONE
> report_variables      NONE
> 
> 
> _______________________________________________
> users mailing list
> [email protected]
> https://gridengine.org/mailman/listinfo/users


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

Reply via email to