[GE users] np_load_avg and thresholds

Don Shesnicky dshesnicky at enqsemi.com
Tue Jul 20 21:04:26 BST 2004


NP = number of processors, load adjusted for processor count... Duh!
 
... and that's why this is off, it thinks this host has 4 processors
even though it only
has 2 Xeons, same as top thinking it has 4.

________________________________

From: Don Shesnicky [mailto:dshesnicky at enqsemi.com] 
Sent: Tuesday, July 20, 2004 3:55 PM
To: users at gridengine.sunsource.net
Subject: RE: [GE users] np_load_avg and thresholds


 
OK I solved the problem with the load and why a certain host was taking
jobs even
though the uptime load was above what I was seeing in both qmon and with
uptime.
 
I started watching the queue via qconf and sure enough when the
np_load_avg hit 1.0+
the host did not accept any more jobs. It had nothing to do with the
uptime load or
the load in qmon.
 
Where are the definitions of np_load_xxx and load_long etc? The np_loads
seem
to have some sort of longer growth/decay factor.
 
$ qconf -se orrin
hostname              orrin.enqsemi.com
load_scaling          NONE
complex_values        NONE
load_values           arch=lx24-x86,num_proc=4,mem_total=3907.785156M, \
 
swap_total=2047.335938M,virtual_total=5955.121094M, \
                      load_avg=2.550000,load_short=0.890000, \
                      load_medium=2.550000,load_long=2.440000, \
                      mem_free=3520.476562M,swap_free=2047.335938M, \
                      virtual_free=5567.812500M,mem_used=387.308594M, \
                      swap_used=0.000000M,virtual_used=387.308594M, \
                      cpu=25.300000,np_load_avg=0.637500, \
<============== Here
                      np_load_short=0.222500,np_load_medium=0.637500, \
                      np_load_long=0.610000
processors            4
user_lists            NONE
xuser_lists           NONE
projects              NONE
xprojects             NONE
usage_scaling         NONE
report_variables      NONE




More information about the gridengine-users mailing list