[GE users] acct_reserved_usage

j_coomer james.coomer at sun.com
Fri Jul 24 20:16:52 BST 2009


Hi all,

I have sge6.2u3. I have set the execd_param in the sge conf to have  ACCT_RESERVED_USAGE=TRUE.

 When I submit a 10 second "sleep" job, the cpu time reported by qacct still reads a small number - rather than "wallclock x slots"

Is this a bug? I would expect the cpu time to equal the wallclock time once I set ACCT_RESERVED_USAGE=TRUE. Or has the behaviour changed from what I remember?

Also, I noticed that the SHARETREE_RESERVED_USAGE has been deprecated. Why? Is there an alternative way to measure "total slots-consumed-time" rather than CPU time? It seems to me that this is the obvious thing to measure in a cluster (you don't care how much CPU people burn - just how much of the cluster they block other users from)

[james at master sge_accounting]$ vi acct.1044
==============================================================
qname        all.q
hostname     comp003
group        james
owner        james
project      NONE
department   defaultdepartment
jobname      serial_job
jobnumber    1044
taskid       undefined
account      sge
priority     0
qsub_time    Fri Jul 24 19:40:40 2009
start_time   Fri Jul 24 19:41:51 2009
end_time     Fri Jul 24 19:42:02 2009
granted_pe   NONE
slots        1
failed       0
exit_status  0
ru_wallclock 11
ru_utime     0.005
ru_stime     0.017
ru_maxrss    0
ru_ixrss     0
ru_ismrss    0
ru_idrss     0
ru_isrss     0
ru_minflt    1461
ru_majflt    0
ru_nswap     0
ru_inblock   0
ru_oublock   0
ru_msgsnd    0
ru_msgrcv    0
ru_nsignals  0
ru_nvcsw     469
ru_nivcsw    7
cpu          0.022
mem          0.000
io           0.000
iow          0.000
maxvmem      103.996M
arid         undefined

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=209383

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



More information about the gridengine-users mailing list