> -----Original Message-----
> From: Testbed Support for GridPP member institutes [mailto:TB-
>
> Ewan, the page that Ian Bird's email pointed to should give you all
the
> information required. If it doesn't then shout.
>
I don't believe it does. The page itself makes no mention of SElinux,
and
simply links to the CIC portal VO cards for details. Following the link
given for ATLAS gives another page that makes no mention of SElinux. I
imagine that it should be near the bottom, under 'Other Requirements',
but it's not.
> I agree that we also need to know how long such things will be
required.
> There is probably an answer for allow_execheap but I doubt you will
get
> a date for 32bit compatibility for example.
>
I have no objection to installing the 32 bit and SL4 compiler version
compatibility RPMs at all, provided that they are available as a
standard
component of SL (note - actual SL, not SLC) so far, they all are. So
long as that doesn't change then the use if these libraries should be
sustainable in the long term. Hacking SElinux about it a different
category
entirely.
> It is always hard to define rules about what will be allowed or not in
> any circumstances. The experiments expect pragmatism and flexibility
> and for the sites truly committed to them they probably get it. As
someone
> else pointed out though, more and more resources are in non-dedicated
> centres.
>
Indeed. The gamble for the VOs is that the more picky they are about the
resources they can run on the fewer resources they will get; the sites
(like Oxford) that are 'truly committed' and own their own hardware can
be flexible (up to a point), but with the best will in the world, not
every cluster is like that.
> I don't think the CERN-based people in the experiments really
> understand this yet. They can boss CERN-IT about so why can't they
boss
> everyone?
>
Yup. I don't know how we can change that culture, but we certainly need
to.
Ewan
|