On Mar 14, 2011, at 19:28 , Rayson Ho wrote:
> By design, the scheduler (the scheduler thread in 6.2) is CPU bound,
> while the qmaster (excluding the scheduler thread) is mostly I/O
> (disk, network, etc) bound. If there is a thread at 100% CPU for 4
> minutes, and only very few random I/O operations, then it is very
> likely that it is the scheduler thread.
>
> Can you turn off qmaster profiling, and turn on scheduler profiling??
> You can enable it by setting scheduler config "PROFILE=TRUE" or
> "PROFILE=1". You will then get the time each stage spends, something
> like:
Thanks, that has helped a bit. I know now that most of the CPU time is spent in
the dispatching stage. However, it is still unclear to me why dispatching
should be such a time-consuming task.
03/15/2011 10:53:56|schedu|master1|P|PROF: job dispatching took 327.370 s (0
fast, 0 fast_soft, 8 pe, 0 pe_soft, 4 res)
03/15/2011 10:53:56|schedu|master1|P|PROF: parallel matching 878
262664 2634 159203 137234 159203 131007
03/15/2011 10:53:56|schedu|master1|P|PROF: sequential matching 0
0 0 0 0 0 0
03/15/2011 10:53:56|schedu|master1|P|PROF: create pending job orders: 0.000 s
03/15/2011 10:53:56|schedu|master1|P|PROF: scheduled in 327.450 (u 337.270 + s
7.960 = 345.230): 0 sequential, 0 parallel, 452 orders, 846 H, 214 Q, 839 QA,
10 J(qw), 431 J(r), 0 J(s), 0 J(h), 0 J(e), 0 J(x), 449 J(all), 57 C, 3 ACL,
149 PE, 12 U, 1 D, 0 PRJ, 1 ST, 0 CKPT, 0 RU, 1 gMes, 0 jMes, 452/3 pre-send,
0/0/0 pe-alg
03/15/2011 10:53:56|schedu|master1|P|PROF: send orders and cleanup took: 0.020
(u 0.020,s 0.000) s
03/15/2011 10:53:56|schedu|master1|P|PROF: schedd run took: 327.630 s (init:
0.000 s, copy: 0.130 s, run:327.470, free: 0.030 s, jobs: 449, categories: 43/0)
A.
--
Ansgar Esztermann
DV-Systemadministration
Max-Planck-Institut für biophysikalische Chemie, Abteilung 105
_______________________________________________
users mailing list
[email protected]
https://gridengine.org/mailman/listinfo/users
