[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [condor-users] Pool node maintenance

Hi Dan,

On Thu, Mar 11, 2004 at 08:36:56AM -0600, Dan Bradley wrote:
> >What I'm looking for is a way to "put the node on hold" so it
> >wouldn't accept any further jobs, without interrupting the one(s)
> >already running on its vm(s). From the manual pages I concluded that
> >condor_off -startd would NOT do the trick, so is there a way to do it,
> >and if not, will it be implemented in a future release at least?
> You can achieve this by temporarily setting the START expression to 
> False on this  host.  That can be done using condor_config_val, but you 
> will probably need to change your security settings to make this 
> possible.  The other way is to simply set START=False in the host's 
> .local config file.  Either way, you'll then need to do condor_reconfig 
> to make the new settings take effect.

The second one worked out of the box (I'm still using a 6.3.1
configuration only patched up a bit to meet the basic requirements of
6.4.7...). Guess I have to look into security more closely before I
upgrade to 6.6.x.

> A new feature currently going into Condor 6.7 allows the startd to 
> publish a promise that it will not kill jobs for at least X amount of 
> run time (where X is some arbitrary ClassAd expression).  Then when 
> preemption happens or you do condor_off, the startd will wait for the 
> current job(s) to finish or for the time limit to run out.

Great! That's something like condor_off -even_more_graceful :-)

Thanks a lot,

 Steffen Grunewald

Steffen Grunewald * * * Merlin cluster admin (http://pandora.aei.mpg.de)
Albert-Einstein-Institut (MPI Gravitationsphysik, http://www.aei.mpg.de)
       Science Park Golm, Am Mühlenberg 1, 14476 Potsdam, Germany
e-mail: steffen.grunewald(*)aei.mpg.de * +49-331-567-{fon:7233,fax:7298}

Condor Support Information:
To Unsubscribe, send mail to majordomo@xxxxxxxxxxx with
unsubscribe condor-users <your_email_address>