[Condor-users] SMP load behavior

[Richard Wackerbarth <richard@xxxxxxxxxx>]

I'm trying to set up a small test pool so that I can get things  
configured to run on a large pool at another institution.
My pool consists of only 2 dual processor PPCs running MAC OSX 10.3  
and 10.4 respectively. The G4 is set to be an "execute" machine and  
the G5 has everything else.

I basically used the configuration from the distribution.

I submitted a few jobs (They each use 100% of a cpu for a couple hours).

Eventually, one of them got assigned to the G4 system.
It ran to completion and the next job took its place as expected.
However, overnight, that system never claimed two jobs at the same time.
Basically, there is nothing else running on the machine.

It appears that the job is running on "vm1" but that it's cpu time is  
getting counted as load on "vm2".
This seems to cause "vm2" to remain in the Owner state.

Have I configured something wrong? Or is this a "bug"?


G5:Warehouse rkw$  /Users/condor/bin/condor_status -state

Name       Cpu Memor LoadAv  KbdIdle    State     StateTime  Acti  
ActvtyTime

vm1@xxxxxx   1   704 0.100   0+13:16:37 Claimed   0+11:03:42 Busy   0 
+00:25:24
vm2@xxxxxx   1   704 1.000   0+13:16:37 Owner     0+10:51:00 Idle   0 
+10:51:00
vm1@xxxxxx   1  2304 0.080   0+00:00:00 Owner     0+14:35:20 Idle   0 
+14:35:20
vm2@xxxxxx   1  2304 0.000   0+00:00:00 Owner     0+03:10:31 Idle   0 
+03:10:31

                     Total Owner Unclaimed Claimed Preempting  
Matched Backfill

                Busy     1     0         0       1          0        
0        0
                Idle     3     3         0       0          0        
0        0

               Total     4     3         0       1          0        
0        0


Thanks for your help.

Richard Wackerbarth
The NFSNet Factoring Group
http://www.nfsnet.org