[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Condor-users] shortening the interval at which the job queue is processed
- Date: Tue, 3 Aug 2010 14:43:28 +0200
- From: Alexandre Fayolle <alexandre.fayolle@xxxxxxxxxx>
- Subject: Re: [Condor-users] shortening the interval at which the job queue is processed
On Tuesday 03 August 2010 14:02:36 Carsten Aulbert wrote:
> Hi Alexandre
> On Tuesday 03 August 2010 13:56:32 Alexandre Fayolle wrote:
> > Some of my users submit quite short lived jobs (completing within a
> > couple of minutes) and complain that they sometimes have to wait for
> > "too long" before the job is started.
> > Can anyone confirm that setting NEGOCIATOR_INTERVAL to e.g. 20 on the
> > central manager will improve the situation? Is there any problem with
> > doing so?
> I think this should help, provided the central manager can cope with the
> extra load.
> If the users put quite a number of jobs into the queue and they are short-
> lived you can try to optimize CLAIM_WORKLIFE on the worker nodes. Using
> this you can bypass the central manager for some time (with the caveat
> that other users might stay in the queue for some time longer as their
> fair share might dictate).
The number of jobs at a given moment is not very high, but they want the
results asap and I typically have enough slots to accomodate the needs.
I'll give it a try and investigate CLAIM_WORKLIFE if necessary.
Alexandre Fayolle LOGILAB, Paris (France)
Formations Python, CubicWeb, Debian : http://www.logilab.fr/formations
Développement logiciel sur mesure : http://www.logilab.fr/services
Informatique scientifique: http://www.logilab.fr/science