[GE users] long running jobs fill all slots

darkstar udo.grabowski at imk.fzk.de
Mon Sep 28 13:05:55 BST 2009


Hello,

the following behaviour looks like a bug to me:

   We've set the share tree policy halflife parameter to 1 hour,
   since shares should be obeyed in time. When sending jobs
   running longer than 1 hour, we get the problem that that
   particular user gets all processing slots after a while
   regardless of the share setting (indeed, even setting his
   shares to 0 does not help!). It looks to me like after
   halflife, still running jobs are not accounted for anymore
   and the system just sends new job, resulting in more and
   more jobs occupying the system. How to get out of this
   situation without setting halflife to a long time ?

   shares are 100% CPU based. Ticket policy is OSF, 0.3 Prio,
   0.1 Urg.,0.6 Ticket, 10000 share tickets,10000 functional
   total share tree and functional 10000.
   auto_user_delete_time 86400
   reprioritize 1
-- 
Dr. Udo Grabowski                           email: udo.grabowski at imk.fzk.de
Institut f. Meteorologie und Klimaforschung ASF,Forschungszentrum Karlsruhe
Postfach 3640, 76021 Karlsruhe, Germany             Tel: (+49) 7247 82-6026
http://www.fzk.de/imk/asf/sat/grabowski/            Fax:         "    -7026

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=219415

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



More information about the gridengine-users mailing list