[GE users] how are functional tickets assigned to jobs?

olson olson at mcs.anl.gov
Fri Nov 7 18:33:48 GMT 2008


I'd forgotten about that, thanks. It didn't seem to say anything about  
the ticket assignments. Here's a snapshot of the issue, qstat -ext  
output (it's really wide, hopefully it'll come thru the list okay).

job-ID  prior   ntckts  name       user         project           
department state cpu        mem     io      tckts ovrts otckt ftckt  
stckt share queue                          slots ja-task-ID
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
<clipped many jobs>
  380015 0.06504 0.00504 mgl_s_3354 mgrastprod   NA                
Dept_mgras qw                                1007     0     0   
1007     0 0.00                                     1
  380016 0.06504 0.00504 mgl_s_3341 mgrastprod   NA                
Dept_mgras qw                                1007     0     0   
1007     0 0.00                                     1
  380017 0.06504 0.00504 mgl_s_3365 mgrastprod   NA                
Dept_mgras qw                                1007     0     0   
1007     0 0.00                                     1
  380018 0.06504 0.00504 mgl_s_3371 mgrastprod   NA                
Dept_mgras qw                                1007     0     0   
1007     0 0.00                                     1
  192459 0.05956 0.01005 ms_1481    mgrastprod   NA                
Dept_mgras qw                                2010     0     0   
2010     0 0.00                                     1 3566-3847:1
  192460 0.05869 0.00917 msvl_1481  mgrastprod   NA                
Dept_mgras qw                                1834     0     0   
1834     0 0.00                                     1 1-28:1
  192461 0.05795 0.00844 msvs_1481  mgrastprod   NA                
Dept_mgras qw                                1687     0     0   
1687     0 0.00                                     1 1-146:1
  192462 0.05732 0.00781 mgg_1481   mgrastprod   NA                
Dept_mgras qw                                1562     0     0   
1562     0 0.00                                     1 1-41:1
  192463 0.05678 0.00727 m16s_1482  mgrastprod   NA                
Dept_mgras qw                                1454     0     0   
1454     0 0.00                                     1 1-38:1
  192464 0.05631 0.00680 ms_1482    mgrastprod   NA                
Dept_mgras qw                                1360     0     0   
1360     0 0.00                                     1 1-3516:1
  192465 0.05590 0.00639 msvl_1482  mgrastprod   NA                
Dept_mgras qw                                1277     0     0   
1277     0 0.00                                     1 1-26:1
  192466 0.05554 0.00602 msvs_1482  mgrastprod   NA                
Dept_mgras qw                                1204     0     0   
1204     0 0.00                                     1 1-133:1
  192467 0.05521 0.00570 mgg_1482   mgrastprod   NA                
Dept_mgras qw                                1139     0     0   
1139     0 0.00                                     1 1-38:1
  334947 0.05504 0.00504 m16s_1580  mgrastprod   NA                
Dept_mgras qw                                1007     0     0   
1007     0 0.00                                     1 1-102:1
  334948 0.05504 0.00504 ms_1580    mgrastprod   NA                
Dept_mgras qw                                1007     0     0   
1007     0 0.00                                     1 1-9731:1
  334949 0.05504 0.00504 msvl_1580  mgrastprod   NA                
Dept_mgras qw                                1007     0     0   
1007     0 0.00                                     1 1-70:1
  334950 0.05504 0.00504 msvs_1580  mgrastprod   NA                
Dept_mgras qw                                1007     0     0   
1007     0 0.00                                     1 1-365:1
<clipped many jobs>

I don't understand why jobs 192459-192467 have more tickets than the  
other queued jobs. They were submitted some time ago, when a different  
assignment of tickets had been made between the departments; is there  
some sticky state left over from that?

algorithm                         default
schedule_interval                 0:0:15
maxujobs                          0
queue_sort_method                 seqno
job_load_adjustments              np_load_avg=0.50
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
compensation_factor               5.000000
weight_user                       0.000000
weight_project                    0.000000
weight_department                 1.000000
weight_job                        0.000000
weight_tickets_functional         1000000
weight_tickets_share              0
share_override_tickets            TRUE
share_functional_shares           TRUE
max_functional_jobs_to_schedule   200
report_pjob_tickets               TRUE
max_pending_tasks_per_job         50
halflife_decay_list               none
policy_hierarchy                  OF
weight_ticket                     1.000000
weight_waiting_time               0.000000
weight_deadline                   3600000.000000
weight_urgency                    0.010000
weight_priority                   0.100000
max_reservation                   0
default_duration                  0:10:00

Here are the dept settings in play currently:

$ qconf -su Dept_mgrast
name    Dept_mgrast
type    DEPT
fshare  1000
oticket 0
entries mgrastprod,mgrasttest
$ qconf -su Dept_rast
name    Dept_rast
type    DEPT
fshare  1000
oticket 0
entries rastprod,rasttest
$ qconf -su defaultdepartment
name    defaultdepartment
type    DEPT
fshare  500
oticket 0
entries NONE

Thanks,
--bob


On Nov 7, 2008, at 11:12 AM, craffi wrote:

> Hi Bob,
>
> There are other mechanisms but one command that may help:
>
> Run "qconf -tsm" as the root user. This will cause a one-time dump of
> scheduling information to a file in $SGE_ROOT/$SGE_CELL/ called
> "schedd_runlog" that may be of assistance.
>
> -Chris
>
> On Nov 5, 2008, at 10:57 PM, olson wrote:
>
>> I've got my cluster set up using functional policy to divide the
>> time on
>> the cluster between the two main applications and all other users,
>> via an
>> allocation to departments.
>>
>> I'm trying to understand the allocation of functional tickets. I  
>> can't
>> figure out why some of the queued jobs have different numbers of
>> tickets
>> and are getting scheduled in a non-obvious order. There aren't
>> override
>> tickets assigned to them.
>>
>> I find information in the docs about how scheduling decisions are  
>> made
>> based on the ticket counts for the jobs, but not how these tickets  
>> are
>> allocated to jobs.
>>
>> thanks for any insights,
>> --bob
>
> ------------------------------------------------------
> http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=88332
>
> To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net 
> ].

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=88335

To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].



More information about the gridengine-users mailing list