[GE users] allow exec host to finish current jobs but not recieve any more queued up jobs

Chris Dagdigian dag at sonsorol.org
Fri Dec 2 03:53:59 GMT 2005


You need to disable a queue (put it into state "d").

You can do this via "qmod -d" -- there is a good manpage ("man qmod")  
for this command as well. ^C

The effect of disabling a queue will get you the behavior you need

  -- running jobs will not be altered or interferred with
  -- no new jobs will enter the queue

You can use this command on any queue instance ("qmod -d  
<queue>@<machinename>"). The command should also allow the wildcard  
('*') character for instances where you want to drain your entire  
cluster in order to perform maintenance (such as "qmod -d '*'").


On Dec 1, 2005, at 8:10 PM, deet wrote:

>   Hello.
>   I'm extremely new to grid engine but I have not been able to find  
> this
> answer in the docs or mailinsg list archives which could be because  
> I dont
> know the terminology for what I'm trying to do.
>    We are using sge-6.0u6 (sles9 amd X86_64).   I have 1 master  
> host and 24
> execution hosts.    I have submitted a few hundred long running  
> jobs into
> the queue via qsub <command to run>.     I need to be able to pause/ 
> disable
> an execution host while allowing the host  to finish it's currently  
> assigned
> jobs but have it not recieve anymore jobs for a period of time until I
> reeanable it.     I've looked through qconf and the closest thing I  
> could
> come up with is delete an exec host (via qconf -de hostname) or via
> modifying the allhosts host group (qconf -mhgrp @allhosts) then  
> removing the
> particular host.    I'm concerned that this is not the correct way to
> achieve what I wish and if this will allow the jobs currently  
> running to
> execute or now.
>    TIA.  Deet.

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