[GE users] Policy question.

reuti reuti at staff.uni-marburg.de
Tue Aug 17 23:56:03 BST 2010


Hi,

Am 17.08.2010 um 23:12 schrieb gutnik:

> I'm using sge 6.2u4.
> 
> I've set up (I think) a Share Tree policy, but it doesn't seem to
> work: jobs get executed in the order they're submitted to the queue.
> I see userA with 50 jobs running for hours. Some time later, userB
> submits a few jobs, but userA's jobs keep getting executed
> while B's wait. As I understand it, userB jobs should be executed
> next, because both show up as "default" users, with equal
> shares.

when you want just a fair-share, then this might be easier to set up:

http://gridengine.info/2006/01/17/easy-setup-of-equal-user-fairshare-policy

but it's not honoring the past usage.


> I have the defaults set for the share tree policy. How would I debug
> this further?
> 
> 
> $ qconf -sstree
> id=0
> name=Root
> type=0
> shares=1
> childnodes=1
> id=1
> name=default
> type=0
> shares=10
> childnodes=NONE

The automatically created user entries will forget the accumulated usage of the past I fear. So two user entries must be created, which won't get deleted at some point ("delete_time 0"). Then these two entries will show up as leafs automatically under the entry "default" (`man share_tree`). Both will get the same share. If you want to have different shares for them, instead of the "default" leaf the two users need be attached directly to the "root" node.

For share-tree setup in a larger scale there are two Howto's:

http://www.sun.com/blueprints/0703/817-3179.pdf
http://www.sun.com/blueprints/1005/819-4325.html

(from http://gridengine.sunsource.net/howto/howto.html)

-- Reuti


> $ qconf -ssconf
> algorithm                         default
> schedule_interval                 0:0:05
> maxujobs                          0
> queue_sort_method                 load
> job_load_adjustments              np_load_avg=1
> 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                          1
> usage_weight_list                 cpu=1.000000,mem=0.000000,io=0.000000
> compensation_factor               5.000000
> weight_user                       0.250000
> weight_project                    0.250000
> weight_department                 0.250000
> weight_job                        0.250000
> weight_tickets_functional         0
> weight_tickets_share              1000
> share_override_tickets            FALSE
> share_functional_shares           FALSE
> max_functional_jobs_to_schedule   200
> report_pjob_tickets               FALSE
> max_pending_tasks_per_job         50
> halflife_decay_list               none
> policy_hierarchy                  S
> weight_ticket                     1.000000
> weight_waiting_time               0.100000
> weight_deadline                   3600000.000000
> weight_urgency                    0.100000
> weight_priority                   0.010000
> max_reservation                   0
> default_duration                  INFINITY
> 
> ------------------------------------------------------
> http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=275033
> 
> To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=275055

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



More information about the gridengine-users mailing list