[GE users] SGE 6.2u4 resource reservation (again)

tkaminski t.kaminski at science-computing.de
Wed Mar 24 13:11:57 GMT 2010


Hi list,

we are using resource reservations for high priority jobs (mostly using
64 slots). These jobs are submitted with -R y, s_rt 02:00:00 and h_rt
02:00:00. The scheduler is configured like this:

------------------------- >8 -------------------------
qconf -ssconf

...
params                            MONITOR=1
max_reservation                   2
default_duration                  9999:00:00
...
------------------------- 8< -------------------------

A look into the schedule file reveals that resource reservation is done.

------------------------- >8 -------------------------
...
28560:1:RESERVING:1305412379:7260:P:starcdhpc9:slots:64.000000
28560:1:RESERVING:1305412379:7260:H:hpc9n007.destr.corpintra.net:slots:8.000000
28560:1:RESERVING:1305412379:7260:Q:debug at hpc9n007.destr.corpintra.net:slots:8.000000
28560:1:RESERVING:1305412379:7260:L:max_slots_on_debug_queue://///:8.000000
28560:1:RESERVING:1305412379:7260:H:hpc9n008.destr.corpintra.net:slots:8.000000
28560:1:RESERVING:1305412379:7260:Q:debug at hpc9n008.destr.corpintra.net:slots:8.000000
28560:1:RESERVING:1305412379:7260:L:max_slots_on_debug_queue://///:8.000000
28560:1:RESERVING:1305412379:7260:H:hpc9n011.destr.corpintra.net:slots:8.000000
28560:1:RESERVING:1305412379:7260:Q:debug at hpc9n011.destr.corpintra.net:slots:8.000000
28560:1:RESERVING:1305412379:7260:L:max_slots_on_debug_queue://///:8.000000
28560:1:RESERVING:1305412379:7260:H:hpc9n012.destr.corpintra.net:slots:8.000000
28560:1:RESERVING:1305412379:7260:Q:debug at hpc9n012.destr.corpintra.net:slots:8.000000
28560:1:RESERVING:1305412379:7260:L:max_slots_on_debug_queue://///:8.000000
28560:1:RESERVING:1305412379:7260:H:hpc9n016.destr.corpintra.net:slots:8.000000
28560:1:RESERVING:1305412379:7260:Q:debug at hpc9n016.destr.corpintra.net:slots:8.000000
28560:1:RESERVING:1305412379:7260:L:max_slots_on_debug_queue://///:8.000000
28560:1:RESERVING:1305412379:7260:H:hpc9n024.destr.corpintra.net:slots:8.000000
28560:1:RESERVING:1305412379:7260:Q:debug at hpc9n024.destr.corpintra.net:slots:8.000000
28560:1:RESERVING:1305412379:7260:L:max_slots_on_debug_queue://///:8.000000
28560:1:RESERVING:1305412379:7260:H:hpc9n026.destr.corpintra.net:slots:8.000000
28560:1:RESERVING:1305412379:7260:Q:debug at hpc9n026.destr.corpintra.net:slots:8.000000
28560:1:RESERVING:1305412379:7260:L:max_slots_on_debug_queue://///:8.000000
28560:1:RESERVING:1305412379:7260:H:hpc9n043.destr.corpintra.net:slots:8.000000
28560:1:RESERVING:1305412379:7260:Q:debug at hpc9n043.destr.corpintra.net:slots:8.000000
28560:1:RESERVING:1305412379:7260:L:max_slots_on_debug_queue://///:8.000000
...
------------------------- 8< -------------------------

All of these slots are already in use. Furthermore, the list of reserved
slots is never changing, even when appropriate slots are getting
available. It seems that free suitable slots are not recognised by the
scheduler. Instead, these slots are used by smaller, low priority jobs.
They are "overtaking" all big, high priority jobs which are starving.

Besides resource reservation a RQS is used:

------------------------- >8 -------------------------
{
   name         max_slots_on_debug_queue
   description  "resource quota for slot usage on debug queue."
   enabled      TRUE
   limit        queues debug to slots=64
}
------------------------- 8< -------------------------

But I think this is not the problem. Does anybody have a hint?


Bye,
Thomas
-- 
Vorstand/Board of Management:
Dr. Bernd Finkbeiner, Dr. Roland Niemeier, 
Dr. Arno Steitz, Dr. Ingrid Zech
Vorsitzender des Aufsichtsrats/
Chairman of the Supervisory Board:
Michel Lepert
Sitz/Registered Office: Tuebingen
Registergericht/Registration Court: Stuttgart
Registernummer/Commercial Register No.: HRB 382196

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=251106

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



More information about the gridengine-users mailing list