LHC Computer Grid - Rollout
> [mailto:[log in to unmask]] On Behalf Of Oxana Smirnova
said:
> Hold on: RB is not needed on every site, VOMS server is not
> needed, even
> a SE does not have to be at every site, and likewise, a CE
> does not have
> to be at a site which is only a storage facility. And so on.
Servers are a different ballgame, I think we're talking about clients
here; you want lcg-cr even if there is no local SE. But it's true that
some pieces may not be needed in particular circumstances. For example,
I suspect there is a voms client RPM which contains, inter alia,
voms-proxy-init and voms-proxy-info. You want both of those on a UI, but
on a WN -info may be useful but -init hopefully not. However, is it
worth the effort of trying to disentangle the two and produce two
separate RPMs purely to save (maybe) a few Mb per installation? And if
you did that you'd probably end up with much longer RPM lists after you
split them.
> Are you aware
> that ATLAS jobs only require ATLAS-x.y.z (and an outbound
> connectivity,
> unfortnately), and run on all possible Linuxes and
> architectures?
Hmm, I'm pretty sure that I saw a mail in the last 24 hours or so saying
that Atlas software is only certified on RH7 and SLC3 (and compatible)
and RH is about to be deprecated ... the jobs may only require the atlas
tag, but the tag is only there if the installation is verified and it
presumably won't verify on a platform where it doesn't work!
Stephen
|