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

Re: [Condor-users] condor_rooster - how does it work



Dan,

I'll give it a go (it may be a while before I get something working)
and I'll let you know how I get on here. The ability to matchmake
with hibernating machines has been on my wish list for a long
time. Currently we have a very simple wakeup cron job which 
assumes that any job can run on any machine at any time. 
This means that our pool is constrained to have machines of
the same/similar spec. Matchmaking would allow us to expand the
pool - if machines are too low spec for users then they just
don't get used. Also I can encourage users with relatively long
jobs to run them out of office hours so we waste less cycles 
( => energy => money ) on badput during the day.

regards,

-ian.

> -----Original Message-----
> From: condor-users-bounces@xxxxxxxxxxx [mailto:condor-users-
> bounces@xxxxxxxxxxx] On Behalf Of Dan Bradley
> Sent: 03 December 2009 15:28
> To: Condor-Users Mail List
> Subject: Re: [Condor-users] condor_rooster - how does it work
> 
> Ian,
> 
> Yes, good point.  You can generate offline ads using condor_advertise.
> Then you could use rooster to manage the wakeup calls and still use some
> other mechanism for hibernation.
> 
> In the ads you give to condor_advertise, just set Offline=True.  Use
> UPDATE_STARTD_AD or INVALIDATE_STARTD_AD in the command argument to
> condor_advertise.
> 
> Let me know if you run into any difficulty.  Once you are happy with how
> it works, this seems like a worthy candidate for the Condor admin how-to
> recipies (hint hint!).
> 
> --Dan
> 
> Smith, Ian wrote:
> > Dan,
> >
> > Only just had a chance to look at this again. That makes sense but I am
> > bit wary of using the Condor power saving at present as we already have
> > a thrid party scheme which has been working pretty well. What I has thinking
> > is - could we populate OFFLINE LOG using say a cron job periodically. We
> > can obtain ClassAds for all machines in the pool and basically just delete
> > those for machines which are powered up to leave the hibernating ones -
> > would this be likely to work.
> >
> > thanks,
> >
> > -ian.
> >
> >
> >> -----Original Message-----
> >> From: condor-users-bounces@xxxxxxxxxxx [mailto:condor-users-
> >> bounces@xxxxxxxxxxx] On Behalf Of Dan Bradley
> >> Sent: 30 November 2009 16:59
> >> To: Condor-Users Mail List
> >> Subject: Re: [Condor-users] condor_rooster - how does it work
> >>
> >> Ian,
> >>
> >> condor_rooster depends on there being offline ads for the hibernating
> >> machines.  The offline ads are created when HIBERNATE becomes true,
> >> assuming your collector has been configured to support offline ads
> >> (configure OFFLINE_LOG).
> >>
> >> Does that help?
> >>
> >> --Dan
> >>
> >> Smith, Ian wrote:
> >>
> >>> Hello All,
> >>>
> >>> As someone who has spent a lot of effort in tying to get our Condor pool to
> >>> work with power saving Windows execute hosts I was very interested in
> >>> using condor_rooster to wake up machines automatically according to
> >>> demand but I cannot for the life of me understand how it works.
> >>>
> >>> Currently the power saving is performed by 3rd party s/w and we do
> >>> not use the hibernating features built into Condor. As far as I can
> >>> tell when the power saving kicks in, the Condor daemons get killed
> >>> and the machine disappears from the pool. I can't work out how
> >>> condor_rooster can work out if these machines are available for
> >>> wakeup when they are no longer publishing ClassAds.  I don't
> >>> see any persistent ClassAds for them - do I need to use the
> >>> HIBERNATE in the execute host config ?
> >>>
> >>> Is there any more documenation available on this - the v. 7.4 manual is a bit
> >>> short on info on this.
> >>>
> >>> thanks - in - advance,
> >>>
> >>> -ian.
> >>>
> >>> --------------------------------------------
> >>> Dr Ian C. Smith,
> >>> e-Science Team,
> >>> The University of Liverpool,
> >>> Computing Services Department
> >>> _______________________________________________
> >>> Condor-users mailing list
> >>> To unsubscribe, send a message to condor-users-request@xxxxxxxxxxx with a
> >>> subject: Unsubscribe
> >>> You can also unsubscribe by visiting
> >>> https://lists.cs.wisc.edu/mailman/listinfo/condor-users
> >>>
> >>> The archives can be found at:
> >>> https://lists.cs.wisc.edu/archive/condor-users/
> >>>
> >>>
> >> _______________________________________________
> >> Condor-users mailing list
> >> To unsubscribe, send a message to condor-users-request@xxxxxxxxxxx with a
> >> subject: Unsubscribe
> >> You can also unsubscribe by visiting
> >> https://lists.cs.wisc.edu/mailman/listinfo/condor-users
> >>
> >> The archives can be found at:
> >> https://lists.cs.wisc.edu/archive/condor-users/
> >>
> > _______________________________________________
> > Condor-users mailing list
> > To unsubscribe, send a message to condor-users-request@xxxxxxxxxxx with a
> > subject: Unsubscribe
> > You can also unsubscribe by visiting
> > https://lists.cs.wisc.edu/mailman/listinfo/condor-users
> >
> > The archives can be found at:
> > https://lists.cs.wisc.edu/archive/condor-users/
> >
> _______________________________________________
> Condor-users mailing list
> To unsubscribe, send a message to condor-users-request@xxxxxxxxxxx with a
> subject: Unsubscribe
> You can also unsubscribe by visiting
> https://lists.cs.wisc.edu/mailman/listinfo/condor-users
> 
> The archives can be found at:
> https://lists.cs.wisc.edu/archive/condor-users/