Ewan, we obviously have a breakdown in UK communications if you are only
now expressing your opinion that it is unreasonable to ask sites to run
anything except out of the box SL5. I have been asking sites via the GDB
monthly since April or May if anyone objects to this allow execheap for
ATLAS? I didn't hear from Oxford. I don't know if the blockage was to
Oxford or from Oxford but we should put something in place to ease the
flow of information.
BTW, what is standard SL5? As well as allow execheap the experiments
have also asked for a variety of 32 bit compatibility libraries and the
gnu version of the compiler, not the SL one. I haven't heard similar
complaints about these.
John
> -----Original Message-----
> From: Testbed Support for GridPP member institutes [mailto:TB-
> [log in to unmask]] On Behalf Of Ewan MacMahon
> Sent: 13 August 2009 11:56
> To: [log in to unmask]
> Subject: Re: Request for sites to move WNs to SL5
>
> > -----Original Message-----
> > From: Testbed Support for GridPP member institutes [mailto:TB-
> >
> > But SELinux is not new, its only that it is turned on in SL5 by
> > default. I don't recall the sites clamouring for the the experiments
> to
> > ensure that their code worked with SELinux on SL4, so I don't think
> we
> > have a very strong case to demand that experiments fix all their
code
> > before we roll out SL5.
> >
> AIUI it was on by default in SL4, and mostly worked fine (up until
> yesterday
> we've had it on on all our SL4 worker node), but SL5 has some changes
> to
> the
> policy that cause problems for some software. It seems to me that when
> an
> experiment is asked to port their code to SL5 then that means SL5 in
> its
> standard configuration, not a bastardised version. We wouldn't accept
a
> VO
> that said "We can run on SL5, assuming you change the C library." (or
> kernel,
> or whatever), and I see no reason that SElinux should be considered
> differently.
>
> Ewan
--
Scanned by iCritical.
|