On 31/07/13 14:00, Daniela Bauer wrote:
>
>
>
> On 31 July 2013 13:39, Ewan MacMahon <[log in to unmask]
> <mailto:[log in to unmask]>> wrote:
>
> > -----Original Message-----
> > From: Testbed Support for GridPP member institutes [mailto:TB-
> <mailto:TB->
> > [log in to unmask] <mailto:[log in to unmask]>] On Behalf
> Of Sam Skipsey
> >
> > Well, ATLAS and CMS do still need a "local configuration area",
> >
> IIRC ATLAS have a cunning plan to not need that at some point in
> the future, but I'm not sure they're quite there yet. For CMS,
> at least the Oxford 'site local' configuration actually lives in
> CVMFS already; we don't have a local NFS area for them at all.
>
> Ewan
>
>
> CMS nearly killed that feature in a recent headless chicken security
> panic ....
>
> On a different note: We left the non-CVMFS software area for CMS
> untouched, i.e. used by SL5 and SL6 and it seems to work just fine. As
> far as I am aware there is no requirement by CMS to make a new software
> area.
> For the experiments that use CVMFS (or try to in our case, sigh), their
> installed software is long gone.
>
>
So, what is best practice?
It would be useful to tell non-LHC VOs what to expect. Even better if
all UK sites to do the same thing.
For CVMFS - one just points at CVMFS. The question is what to do for
others.
My current view is that having a fresh start allows (indeed forces) them
to compile their software for SL6 - and that this is a good thing.
Therefore we should do that (making sure that software tags reflect
that). If VOs want different treatment, they can talk to sites directly.
Would anyone object to me making that statement?
Chris
|