[GE users] A little help with SGE commands

Chris Dagdigian dag at sonsorol.org
Thu Sep 20 17:03:42 BST 2007


    [ The following text is in the "ISO-8859-1" character set. ]
    [ Your display is set for the "ISO-8859-10" character set.  ]
    [ Some special characters may be displayed incorrectly. ]

Hi Luis,

My replies are inline with your message ...

First off, this is your best source for documentation:

http://docs.sun.com/app/docs/coll/1017.3?l=en&q=N1GE

On Sep 20, 2007, at 11:53 AM, Luís Miguel Silva wrote:

> Hello everybody,
>
> I''ve just started messing around with SGE and i was wondering if  
> anybody could
> help me find out how to:
> - check existing nodes on cluster

Run the commands "qstat -f" and "qhost"

Read the manpages for "qstat" and "qhost" to learn how they work

> - check node state

Use the "qstat" program

> - how to add a compute node to the cluster

Run the script "./install_execd" or "./inst_sge -x" from within  
$SGE_ROOT while logged into the compute nodes.

For info on some of the prerequisites and dependencies, this blog  
article may be of use:
http://gridengine.info/articles/2005/09/29/things-to-think-about- 
before-installing


> - how to ask more then one compute node during job submission

More than one node only works with parallel jobs. You'll need to  
create a parallel environment (PE). The SGE documentation on parallel  
environments will help here.

> - is there any way to run a script after a job is finished /  
> canceled / failed ?

These are called epilog scripts; there are a number of ways to do  
this depending on your specific needs.

>
> After i installed the SGE service, i think i added my two compute  
> nodes (the
> headnode + a dedicated compute node) but i don''t know how to check  
> if i
> configured them correctly.
>

"qstat -f" is your friend

> Every job i submit is executed on my compute node so im guessing my  
> headnode is
> not properly configured as a compute node.
>

This could be true or your head node may be "too busy" as measured by  
various load indexes measured by SGE. It could be normal behavior on  
your system. "qstat -f" will let you know of any real problems.

If all else fails you can explicitly request your head node run a job:

   qsub -q all.q at my-head-node ./my-test-script.sh

   qrsh -q all.q at my-head-node "date; hostname; uptime"


> Could anybody point me to some "cheat sheets" with simple / direct  
> examples on
> SGE''s different commands / syntaxes?



---------------------------------------------------------------------
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