[GE users] calendar and no h_rt request

reuti reuti at staff.uni-marburg.de
Wed Oct 20 15:49:16 BST 2010


Hi,

Am 20.10.2010 um 16:15 schrieb murple:

> to prepare for a planned downtime I created a calendar a added it to the queue.
> Now when I submit a job without explicitly requesting h_rt I will not be started 
> "due to a reservation" despite that the queues h_rt limit ends before the downtime.
> 
> If I explicitely request h_rt (with the same value as queue limit) it works.

yep, the "default_duration" will be applied when the reservation is checked, which is "infinity". You could lower it to a more feasible value (up to h_rt), which would also avoid wrong backfilling (SGE judges "infinity" being smaller than "infinity", and so jobs might slip in for backfilling although they shouldn't. I set it to 9999:99:99 to avoid this side effect).

-- Reuti


> This is 6.2 (u0)
> 
> here my setup:
> 
> kuntzagk at login2:~> qconf -sq test
> qname                 test
> hostlist              @nodes @nodes_v2
> seq_no                20
> load_thresholds       np_load_avg=1.75
> suspend_thresholds    NONE
> nsuspend              1
> suspend_interval      00:05:00
> priority              0
> min_cpu_interval      00:05:00
> processors            UNDEFINED
> qtype                 BATCH
> ckpt_list             NONE
> pe_list               make orte smp
> rerun                 TRUE
> slots                 4,[@fastq_nodes=2],[@nodes_v2=8],[node001=8]
> tmpdir                /tmp
> shell                 /bin/csh
> prolog                NONE
> epilog                NONE
> shell_start_mode      unix_behavior
> starter_method        NONE
> suspend_method        NONE
> resume_method         NONE
> terminate_method      NONE
> notify                00:00:60
> owner_list            NONE
> user_lists            root
> xuser_lists           NONE
> subordinate_list      longrun
> complex_values        NONE
> projects              NONE
> xprojects             NONE
> calendar              downtime_04_11_2010
> initial_state         default
> s_rt                  96:00:00
> h_rt                  96:00:00
> s_cpu                 INFINITY
> h_cpu                 INFINITY
> s_fsize               INFINITY
> h_fsize               INFINITY
> s_data                INFINITY
> h_data                INFINITY
> s_stack               INFINITY
> h_stack               256M
> s_core                INFINITY
> h_core                INFINITY
> s_rss                 INFINITY
> h_rss                 INFINITY
> s_vmem                INFINITY
> h_vmem                8G
> 
> kuntzagk at login2:~> qconf -scal downtime_04_11_2010
> calendar_name    downtime_04_11_2010
> year             4.11.2010-5.11.2010=17:30-8:00=off
> week             NONE
> 
> kuntzagk at login2:~> qsub -q test  /opt/sge/examples/jobs/simple.sh
> kuntzagk at login2:~> qstat -j
> ..
> (-l NONE) cannot run at host "node055" because for default request it offers 
> only hc:h_vmem=536870912.000000
> 
> kuntzagk at login2:~> qalter -l h_rt=96:00:00 /opt/sge/examples/jobs/simple.sh
> 
> .. works
> 
> regards, Andreas
> 
> ------------------------------------------------------
> http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=288639
> 
> To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=288641

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



More information about the gridengine-users mailing list