Hi,
Right, but Geant4 at this point is running as dteam for some reasons:
1. They run on Grid basically twice per year, at the same time they
produce new tags they test them in LCG2 but this is basically some weeks
in April or May and some weeks in November or December. Since they are
running since last December and at the beginning they were just performing
some test to see the performance of Grid in their analysis, we propose
them to run as dteam. Geant4 is in principle a small collaboration and
just two persons make the whole production in Grid.
2. Geant4 is a software already tested in many platforms and nodes (they
are producing simulations for more than 10 years). Since their software is
very stable we have considered the idea of using some parts of their
software and procedures as tests for our own system. This should be done
as dteam since we are testing our own sites.
3. Geant4 is a toolkit used by many experiments (ATLAS, CMS, Babar, some
Biomed communities, LHCb). So if all these experiments are bringing Geant4
as part of their software installation it results at the end that the
experiment software area of ATLAS, CMS, Babar... have Geant4 installed. We
could installed more than one Geant4 version in the dteam area of all
sites where the rest of VOs make production and they could read the
software from that area. In somehow it to repeat the situation of the
experiments running Geant outside Grid. Geant4 have the software already
compiled in a certain area of /afs at cern. Normally the Geant4 users do
not install the whole toolkit in their own machines but link the
installation of the afs area. Something similar could be applied to the
Grid if this software is installed in the dteam area of the WNs and the
rest of experiments just link to the corresponding area.
Cheers,
Patricia
On Tue, 17 May 2005, Kostas Koumantaros wrote:
> Hi,
>
> Correct me if I'm wrong DTEAM VO should not be used by experiments as
> its purpose is mainly deployment and administration.
> Doesn't GEANT4 belong to a specific VO ?
>
> Thanks,
>
> Kostas
>
>
>
> Patricia MENDEZ LORENZO wrote:
>
> >Hello,
> >
> >
> >
> >> I have a very stupid question: how experiment software managers are 'triggered'
> >>to install software on a given site? The matter is, that once we've installed
> >>
> >>
> >
> >This depends on each VO. Each VO can decided how many of their members can
> >install the software in the delivered area of the WNs at each site. For
> >example in the case of DTEAM we can be around 3 persons but this changes
> >for another VO.
> >
> >
> >
> >>LCG-2.4.0 on our site I was contacted by DTEAM SGM and she was installed GEANT
> >>on our site. There were no requests from my side to install that software, she
> >>just mailed me that she is installing the software. This gave me the sense, that
> >>
> >>
> >
> >The request did not come from the site but from the experiment. In this
> >case it was a Geant4 requirement and at this point they are running as
> >dteam. Geant4 wanted to make a huge production and before they have to
> >split their software in the largest possible number of sites. So its
> >software has been installed in more than 20 sites in the DTEAM Experiment
> >software area of the WNs.
> >
> >
> >
> >>experiment SGM's are just watching for new or reinstalled sites and then, when
> >>all tests are fine, they are installing their software on that site. (If it
> >>
> >>
> >
> >Not exactly, it depends on the status of their production. If a certain VO
> >is not in production they will not update the software. What I made after
> >many sites migrated to LCG2.4.0 was to test the experiment software area
> >in the WNs and the /opt/edg/var/<vo>/<vo>.list files of the CEs. Once
> >these areas were fine I published a tag in the corresponding site for my
> >own information (to know the site was tested and was working fine. The tag
> >was called VO-dteam-testingthesite1). If any problem in this sense was
> >found I sent an email to the corresponding site.
> >
> >
> >
> >>is permitted by site's configuration). But after a week of waiting I'm not
> >>sure that it it true. Can someone explain it to me, how to trigger software
> >>installation and validation of a site by VO SGMs?
> >>
> >>
> >
> >The installation and validation depends again on each VO. They decided
> >how to make the installation (maybe it's just a tar file, but maybe they
> >need more complicate installation tools) and they decide when and how this
> >installation is validated (in the case of Geant4 once the file has been
> >untar they produce a very small ntuple and at this point they validate the
> >installation, again this depends on the experiment). What we propose them
> >are different tools to perform the installation, the validation and the
> >tag publication but they decided how to use these tools depending on their
> >needs.
> >What we do is to check together with the sites the status of the software
> >area, shared areas among WNs, installation and configuration of Tank&Spark
> >and the /opt/edg/var/<vo>/<vo>.list area of the CEs.
> >
> >Cheers,
> >Patricia
> >
> >
> >
> >
> >
> >
> >> Thanks.
> >>
> >>
> >>
> >
> >
> >
>
--
EUROPEAN LABORATORY FOR PARTICLE PHYSICS -- CERN
Patricia Mendez Lorenzo
IT/GD Division
Building: 28 Office: R-019
Phone: +41 22 767 1925
Fax: +41 22 767 7155
Email: [log in to unmask]
|