Custom Query (431 matches)

Filters
 
Or
 
  
 
Columns

Show under each result:


Results (85 - 87 of 431)

Ticket Resolution Summary Owner Reporter
#1457 wontfix qstat reports urgency slots rather than real/range value wish
Description

If you set the urgency slots of a pe to a constant then a normal 'no flags' qstat will report that as the number of slots a job requesting the job requires: wish@iseisd-ws113:~$ qstat|tail -n +3 |awk '{print $1}'|xargs -n 1 qstat -j '{}' |grep 'parallel';qstat parallel environment: smp range: 2 parallel environment: smp range: 2 parallel environment: smp range: 4 parallel environment: smp range: 4 parallel environment: smp range: 2-4 parallel environment: smp range: 2,3 job-ID prior name user state submit/start at queue slots ja-task-ID


10 0.42045 fellow wish r 03/26/2013 09:13:40 all.q@… 2 13 0.41629 fellow wish r 03/26/2013 09:21:55 all.q@… 2

9 0.75000 fellow wish qw 03/26/2013 09:08:27 4

11 0.74968 fellow wish qw 03/26/2013 09:08:34 4 12 0.25536 fellow wish qw 03/26/2013 09:08:40 1 14 0.25000 fellow wish qw 03/26/2013 09:10:37 1

#1456 fixed min_uid and min_gid apply to prefix users of prolog Dave Love <d.love@…> wish
Description

Setting min_uid and min_gid prevents sge_execd from running a prolog as a user with a uid or gid below this value using the user@ prefix. This can be worked around (eg by running it as a special user with uid/gid above the threshold and using sudo to switch to the desired user) it is unexpected and adds a difficulty for no real security gain. The effect of this failure to run the prolog seems to be to push the job back on to the queue.

The same problem probably applies to the epilog.

#1455 fixed Wrong load for cluster queue with 0 slots Dave Love <d.love@…> njoly
Description

When a cluster queue is that has hosts but is set to have 0 slots the rported load for that queue is wrong:

njoly@lanfeust [~]> qstat -g c CLUSTER QUEUE CQLOAD USED RES AVAIL TOTAL aoACDS cdsuE


empty nan 0 0 0 0 0 0

The reported "nan" value results in a division by zero ...

The provided patch do fix it by marking load as unavailable if there is no slots in the cluster queue.

njoly@lanfeust [~]> qstat -g c CLUSTER QUEUE CQLOAD USED RES AVAIL TOTAL aoACDS cdsuE


empty -NA- 0 0 0 0 0 0

Note: See TracQuery for help on using queries.