[Mauiusers] policy confusion

Paul Van Allsburg vanallsburg at hope.edu
Thu Nov 9 11:19:34 MST 2006

Lennart Karlsson wrote:

>My advice is that you try some of the informational commands of
>Maui, like
>	showq
>	mdiag -q
>	checkjob 5846
>	showstart 5846
>and perhaps also
>	checkjob -v 5846
>These will possibly tell you quite a lot about what resources
>there are and why your job are not able to run on them.
>Best regards,
>-- Lennart Karlsson <Lennart.Karlsson at nsc.liu.se>
>   National Supercomputer Centre in Linkoping, Sweden
>   http://www.nsc.liu.se
>Paul Van Allsburg wrote:
>>I have what seems to be a simple policy but my job is stuck in the queue 
>>and I don't know why.  The cluster is 16 nodes/32processors, I have 4 
>>queues, 'normal' is the default.   The is the cluster current status:
>>Job id           Name             User             Time Use S Queue
>>---------------- ---------------- ---------------- -------- - -----
>>5805.curie       ...o1-7imp-md123 hinkle           13:37:24 R 
>>5836.curie       ...o1-2imp-md125 hinkle                  0 Q 
>>5837.curie       ...o1-9imp-md136 hinkle                  0 Q 
>>5846.curie       cpuburn          vanallp                 0 Q normal
>>I have Hinkle limited to  4 processors,  job 5805 is using all 4.  I 
>>submitted cpuburn to a single node but it's not running. 
>>My maui.cfg is:
>># maui.cfg 3.2p8
>>RMPOLLINTERVAL        00:02:00
>>SERVERPORT            42559
>>LOGFILE               maui.log
>>LOGFILEMAXSIZE        10000000
>>LOGLEVEL              3
>>CREDWEIGHT            1
>>USERWEIGHT            1
>>GROUPWEIGHT           1
>>CLASSWEIGHT           1
>>USERCFG[vanallp]      MAXNODE=2
>>USERCFG[hinkle]       MAXPROC=4
>>USERCFG[webmo]        MAXNODE=4 PRIORITY=100000
>># these are the 4 queues
>>CLASSCFG[webmoq]      PRIORITY=1000000
>>CLASSCFG[normal]      MAXNODE=14
>>CLASSCFG[debug]       MAXNODE=15
>>CLASSCFG[admin]       MAXNODE=16
>>XFACTOR               1
>># this parm gives short wall clock jobs priority
>># limited to 1 day... see in Maui admin guide:)
>>#                     one day!    
>>XFMINWCLIMIT          1440
>>Am I missing the obvious? 
>>Paul Van Allsburg
I think I'm a little more confused... I did a
checkjob 5846
and it immediately returned with a State: of "Running" on node 8. 
I qsub'ed another, and it immediately started.  I qsub'ed 14 more and 
they all ran. 
It seems the MAXNODE= has no effect on the scheduler in my 
configuration.  When I set
MAXPROC=  the scheduler will correctly hold jobs based on that 
setting.   Where did I
go wrong?     

More information about the mauiusers mailing list