[GE users] [GE dev] Load sensor scripts

Reuti reuti at staff.uni-marburg.de
Tue Aug 17 13:13:59 BST 2004


The usual place is any of the messages files of SGE in 
$SGE_ROOT/default/spool/qmaster or the nodes directories.
Did you tried to start the load sensor script interactive, so that it 
gives the correct output after each <return>?


Dev wrote:

>Didnt work for me. I am sure I am missing something
>here. Have to figure it out. I do remember having
>retrieved the value using qstat -F a few days back but
>now it dosnt show up. maybe some script problem. Is
>there any way of checking in some log whether the
>script was actually executed on the exec host?
>
>
>--- Reuti <reuti at staff.uni-marburg.de> wrote:
>
>  
>
>>Hi,
>>
>>Try to change the line:
>>
>>if [ $input = "quit" ]; then
>>
>>to
>>
>>if [ "$input" = "quit" ]; then
>>
>>It will take some time, until the value will show
>>up.
>>
>>
>>Maybe the Administration Guide should be updated as
>>well (if it's this 
>>way working for you).
>>
>>Cheers - Reuti
>>
>>
>>Dev wrote:
>>
>>    
>>
>>>HI,
>>>
>>>  I have tried the suggestion but I still cant
>>>      
>>>
>>find
>>    
>>
>>>my load value when I do qstat -F.
>>>
>>>This is what I have done
>>>
>>>created a shell script
>>>
>>>####!/bin/sh
>>>
>>>myhost=`uname -n`
>>>
>>>while [ 1 ]; do
>>>    # wait for input
>>>    read input
>>>    result=$?
>>>    if [ $result != 0 ]; then
>>>         exit 1
>>>    fi
>>>    if [ $input = "quit" ]; then
>>>         exit 0
>>>    fi
>>>    #send users logged in
>>>    logins=`who -q | tail -1 | cut -d = -f 2`
>>>    echo begin
>>>    echo "$myhost:ge_load:$logins"
>>>    echo end
>>>done
>>>
>>># we never get here
>>>
>>>exit 0
>>>
>>>The script is supposed to send the number of logged
>>>      
>>>
>>in
>>    
>>
>>>users.
>>>
>>>I edited the cluster configuration select global
>>>      
>>>
>>and
>>    
>>
>>>do a modify and enter the path of the script in the
>>>load sensor box.
>>>
>>>I then goto complex configuration and create a
>>>      
>>>
>>complex
>>    
>>
>>>attribute ge_load which is as follows
>>>#name               shortcut        type       
>>>      
>>>
>>relop
>>    
>>
>>>requestable consumable default  urgency
>>>      
>>>
>>#---------------------------------------------------------------------------------------------
>>    
>>
>>>arch                a               RESTRING    == 
>>>      
>>>
>> 
>>    
>>
>>>YES         NO         NONE     0
>>>calendar            c               RESTRING    == 
>>>      
>>>
>> 
>>    
>>
>>>YES         NO         NONE     0
>>>cpu                 cpu             DOUBLE      >= 
>>>      
>>>
>> 
>>    
>>
>>>YES         NO         0        0
>>>ge_load             ge_load         INT         >= 
>>>      
>>>
>> 
>>    
>>
>>>NO          NO         0        0
>>>....rest omitted 
>>>
>>>Now I try to do qstat -F and my ge_load value
>>>      
>>>
>>doesnt
>>    
>>
>>>exist there.
>>>
>>>I have also tried doing qconf -se "exec host name"
>>>      
>>>
>>.
>>    
>>
>>>This displays the default load values but not my
>>>custom load value. 
>>>
>>>even if qstat -F does display the load value that
>>>      
>>>
>>will
>>    
>>
>>>be only for that particular exec host where we run
>>>qstat -F. But if I want to get the individual
>>>      
>>>
>>values
>>    
>>
>>>sent by each sge_execd to qmaster how will I get
>>>that?. Do I need to use GDI or something to access
>>>      
>>>
>>it
>>    
>>
>>>directly from the qmaster??
>>>
>>>Please correct me if my understanding somewhere has
>>>been wrong.
>>>
>>>cheers,
>>>
>>>Dev
>>>
>>>
>>>--- "Olesen, Mark" <Mark.Olesen at arvinmeritor.com>
>>>wrote:
>>>
>>> 
>>>
>>>      
>>>
>>>>>  When we add load sensor scripts to the global
>>>>>cluster configuration and create a complex
>>>>>     
>>>>>
>>>>>          
>>>>>
>>>>attribute
>>>>   
>>>>
>>>>        
>>>>
>>>>>for it as well, How can we retrieve the value of
>>>>>     
>>>>>
>>>>>          
>>>>>
>>>>this
>>>>   
>>>>
>>>>        
>>>>
>>>>>attribute? Moreover when the script is added to
>>>>>     
>>>>>
>>>>>          
>>>>>
>>>>the
>>>>   
>>>>
>>>>        
>>>>
>>>>>global cluster configuration where will the
>>>>>          
>>>>>
>>script
>>    
>>
>>>>>     
>>>>>
>>>>>          
>>>>>
>>>>be
>>>>   
>>>>
>>>>        
>>>>
>>>>>run? WIll it be on the sge_master machine? or can
>>>>>     
>>>>>
>>>>>          
>>>>>
>>>>the
>>>>   
>>>>
>>>>        
>>>>
>>>>>sge_master instruct all the sge_execs to run the
>>>>>script and use the data returned by them to
>>>>>     
>>>>>
>>>>>          
>>>>>
>>>>generate
>>>>   
>>>>
>>>>        
>>>>
>>>>>the final effective load value?
>>>>>     
>>>>>
>>>>>          
>>>>>
>>>>I've redirected the response to the USERS mailing
>>>>lists, since the question
>>>>is not really a development issue.
>>>>
>>>>
>>>>The load sensor in the global configuration is run
>>>>by *all* execd's.
>>>>If you wish to run the load sensor on the qmaster,
>>>>you need to install a
>>>>queue instance on the qmaster machine. You could,
>>>>for example, have a queue
>>>>instance with zero slots on the qmaster.  This
>>>>        
>>>>
>>would
>>    
>>
>>>>ensure that no jobs run
>>>>on the qmaster, but a loadsensor will.
>>>>
>>>>How the values are reported is documented in
>>>>sge_exed(8).
>>>>
>>>>Within the loadsensor, you can differentiate
>>>>        
>>>>
>>between
>>    
>>
>>>>qmaster and normal
>>>>execd with something like the following:
>>>>
>>>>  if [ "$UQHOST" = "$qmaster" ]; then
>>>>     issue_some_global_report
>>>>  fi
>>>>
>>>>You can query the return values via 'qstat -F' ...
>>>>provided, of course, that
>>>>you have defined the complexes!
>>>>
>>>>/mark
>>>>
>>>>
>>>>Dr. Mark Olesen
>>>>Thermofluid Dynamics Analyst
>>>>ArvinMeritor Light Vehicle Systems
>>>>Zeuna Staerker GmbH & Co. KG
>>>>        
>>>>
>=== message truncated ===
>
>
>
>		
>__________________________________
>Do you Yahoo!?
>New and Improved Yahoo! Mail - Send 10MB messages!
>http://promotions.yahoo.com/new_mail 
>
>---------------------------------------------------------------------
>To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
>For additional commands, e-mail: users-help at gridengine.sunsource.net
>  
>


---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list