My interpretation is that CREAM is ready and used by Alice in production
but it will run alongside the lcg-CE until:-
a) the ICE component in WMS to allow submission via WMS works reliably.
b) CREAM is demonstrated to be at least as good (performant, reliable,
??) as LCG-CE.
The latter needs sites and VOs to try it which is why sites are asked to
run one.
I think the likelihood of a decision to drop lcg-CE before data taking
is low. If CREAM was complete rubbish then there would be pressure to
port the lcg-CE to SL5 and I don't see that. I'm not aware that the
other LHC experiments are planning to use new functionality in CREAM so
the migration should be smooth. Once it has proven itself sites can drop
their lcg-CEs to move away from SL4.
Of course I may be wrong and Panda or Ganga may use some new feature
like JSDL and there will then be pressure on everyone to run CREAM and
drop lcg-CE. Not soon though.
John
> -----Original Message-----
> From: Testbed Support for GridPP member institutes [mailto:TB-
> [log in to unmask]] On Behalf Of Christopher J. Walker
> Sent: 20 August 2009 09:05
> To: [log in to unmask]
> Subject: Re: WLCG migration of WNs to SL5
>
> Gordon, John (STFC,RAL,ESC) wrote:
> > I can confirm Dug's comments from the wider GDB perspective. CREAM
> seems
> > to work well but no WMS job submission yet so stay with lcg-CE.
> >
>
> Do you expect that CREAM will be used when data taking starts?
>
> Or to put it another way, do I build an lcg-CE on older hardware for
> testing the SL5 worker nodes - with the expectation that I'll use
CREAM
> on a newer machine in the end, or should I put the lcg-CE on the newer
> machine.
>
> Chris
>
> >
> >
> > More volunteers to try CREAM welcome though.
> >
> >
> >
> > JOhn
> >
> >
> >
> > *From:* Testbed Support for GridPP member institutes
> > [mailto:[log in to unmask]] *On Behalf Of *Douglas McNab
> > *Sent:* 13 August 2009 11:09
> > *To:* [log in to unmask]
> > *Subject:* Re: WLCG migration of WNs to SL5
> >
> >
> >
> > Hi,
> >
> > In the UK, RAL & Glasgow have CREAM CE's in 'production', well in
the
> > GOCDB anyway.
> >
> > From Glasgow perspective it is used for a set of local users with
> > direct submission as the WMS ICE component required for submission
is
> > still broken. We use it for our local optics VO as CREAM has a much
> > easier way of adding functionality to the job manager via a plugin
> > mechanism consisting of bash scripts. We use it for adding a
> consumable
> > resource when arbitarty JDL is specified. Something that on an lcg-
> CE
> > would require 'tweaking' the job manager itself.
> >
> > Observations on cream: Direct Submission is fast and JDL based, the
> > same WMS JDL should work with direct submission to CREAM. Output has
> to
> > be returned via gridftp automatically. So no need for users to
> retrieve
> > output or forget to retrieve it but the requirement for a gridftp
> > server. Easy to tweak job manager. Not really subjected it to
heavy
> > load so do not know how it would handle large numbers of jobs.
> >
> > Atlas cannot submit via CREAM as there are issues with condor_g
> although
> > a patch for this does exist I believe.
> > ALICE are using CREAM now, not so sure about LHCb or CMS.
> >
> > However, as Jeremy stated it has still not met the criteria to
> replace
> > the lcg-CE. So I would use another lcg-CE for now.
> >
> > Cheers,
> >
> > Dug
> >
> > 2009/8/13 Coles, Jeremy (STFC,RAL,ESC) <[log in to unmask]
> > <mailto:[log in to unmask]>>
> >
> > Hi Winnie
> >
> > I am replying on TB-SUPPORT not the GridPP-USERS list as I think
this
> is
> > a question more relevant to other sysadmins than UK users.
> >
> > The CREAM CE has still not met the criteria put in place to indicate
> > when it is accepted
> > (https://twiki.cern.ch/twiki/bin/view/LCG/LCGCEtoCREAMCETransition )
> to
> > replace the LCG CE. Therefore it is not recommended.
> >
> > A couple UK sites have played with the basic CREAM CE - perhaps
those
> > involved could respond to the list with their observations?
> >
> > Thanks,
> > Jeremy
> >
> >
> >
> > -----Original Message-----
> > From: Discussion list for GridPP Users
> > [mailto:[log in to unmask]
> > <mailto:[log in to unmask]>] On Behalf Of Winnie Lacesso
> > Sent: 13 August 2009 09:26
> > To: [log in to unmask] <mailto:[log in to unmask]>
> > Subject: Re: WLCG migration of WNs to SL5
> >
> > Dear Doctor,
> >
> > Bristol's got some new SL5 WN not yet in production - got to build a
> > new CE to drive them, question is should it be CREAM-CE or lcg-CE?
> > A while back the concensus seemed to be CREAM-CE was problematic or
> > just didn't work. Has it improved?
> > Any UK sites had experience with or using CREAM-CE?
> > Is it stable/fine or still not yet ready for prime time?
> >
> > ldapsearch AFAICS shows only 2 sites using it:
> > GlueForeignKey: GlueClusterUniqueID=cream-ce01.marie.hellasgrid.gr
> > <http://cream-ce01.marie.hellasgrid.gr>
> > GlueForeignKey: GlueClusterUniqueID=cream-ce.pdc.kth.se
> > <http://cream-ce.pdc.kth.se>
> >
> > wl / got -5 min to waste time if CREAM-CE is borken
> > --
> > Scanned by iCritical.
> >
> >
> >
> >
> > --
> > ScotGrid, Room 481, Kelvin Building, University of Glasgow
> > tel: +44(0)141 330 6439
> >
> >
> > --
> > Scanned by iCritical.
> >
> >
--
Scanned by iCritical.
|