[gridengine users] problem with reservation. It never gets active...

Reuti reuti at staff.uni-marburg.de
Sat Jul 13 16:37:36 UTC 2013


Am 10.07.2013 um 15:42 schrieb Pablo Escobar:

> Hi,
> 
> I have a problem where reservation doesn't seem to work. It was
> working previously but now it stopped working and I don't know why. I
> think is it related to jobs walltime (h_rt) but not sure.
> 
> I have a few queues and use "seqno" in the queues. I also have two
> "special" queues which require a complex value (long=true or
> fast=true). I also use h_vmem reservation in every queue
> 
> Here is my config:
> 
> [root][sgemaster][~]# qconf -ssconf|grep -i reser
> max_reservation                   30
> 
> [root][sgemaster][~]# qconf -ssconf|grep -i reser
> max_reservation                   30
> 
> [root][sgemaster][~]# qconf -ssconf|grep -i dura
> default_duration                  INFINITY

Don't use INFINITY here as SGE will judge INFINITY being smaller than INFINITY and other jobs may always slip in as they fit into the remaining timeframe. If you always request h_rt of course, than you are safe as this value won't be used at all.


> [root][sgemaster][~]# for q in `qconf -sql`; do echo $q ; qconf -sq $q
> |egrep "h_rt|s_rt|complex";echo -e "\n"; done
> 
> 12cores128G.q
> complex_values        NONE
> s_rt                  INFINITY
> h_rt                  16:00:00

The queue limits are not taken into account for reservation (it's an RFE, as it makes of course sense to use the maximum absolute runtime in this queue as default).

-- Reuti


> 12cores96G.q
> complex_values        NONE
> s_rt                  INFINITY
> h_rt                  16:00:00
> 
> 
> 8cores30G.q
> complex_values        NONE
> s_rt                  INFINITY
> h_rt                  20:00:00
> 
> 
> 8cores48G.q
> complex_values        NONE
> s_rt                  INFINITY
> h_rt                  16:00:00
> 
> 
> clinic.q
> complex_values        clinicq=true
> s_rt                  INFINITY
> h_rt                  INFINITY
> 
> 
> fast.q
> complex_values        fast=TRUE
> s_rt                  INFINITY
> h_rt                  INFINITY
> 
> 
> gpus-all.q
> complex_values        gpusallq=true
> s_rt                  INFINITY
> h_rt                  24:00:00
> 
> 
> gpus.q
> complex_values        NONE
> s_rt                  INFINITY
> h_rt                  16:00:00
> 
> 
> long.q
> complex_values        long=true
> s_rt                  INFINITY
> h_rt                  INFINITY
> 
> [root][sgemaster][~]# qconf -sc |egrep "^#|slots|fast|long|h_vmem" |grep -v load
> #name               shortcut   type        relop requestable
> consumable default  urgency
> #----------------------------------------------------------------------------------------
> fast                f          BOOL        ==    FORCED      NO
> FALSE    100
> h_vmem              h_vmem     MEMORY      <=    YES         YES
> 0        0
> long                long       BOOL        ==    FORCED      NO
> FALSE    100
> slots               s          INT         <=    YES         YES
> 1        1000
> 
> 
> I have checked many times that flag "-R y" is active in a few jobs and
> enabled "monitor=true" in my scheduler config for debugging but I get
> no information about reserveration in the "schedule" file. It is like
> reservation never gets active :?
> 
> 
> [root][sgemaster][~]# grep -i reser $SGE_ROOT/$SGE_CELL/common/schedule
> [root][sgemaster][~]#
> 
> 
> any clue about what else I could check?
> 
> thanks in advance for any help.
> regards,
> Pablo.
> _______________________________________________
> users mailing list
> users at gridengine.org
> https://gridengine.org/mailman/listinfo/users





More information about the users mailing list