The point for me was not so much whether it is a standard WN OS or a
standard CE OS ... just that it better be some OS for which I already
have an automatic install setup. For example if someone says RHEL
compatible I am happy, if someone says SLC 3.0.4 I am not because we
don't have SLC here at all.
JT
Henry Nebrensky wrote:
>>Åke Sandgren wrote:
>>
>>>The site should decide what OS they want to run and the VO should be
>>>made to adapt to this. The site may for instance have security rules in
>>>place that says they can't run just any OS that the VO wants (or anyone
>>>of the EGEE/LCG/EDG supported ones) or they might even be running a
>>>"non-supported" hardware like RS6K machines with AIX...
>
> ...
>
> I don't really understand this: the LCG middleware hasn't been ported much
> beyond Fedora has it - certainly not (e.g.) AIX?
>
> Whatever the WNs use, I would expect that the service nodes at a site
> would be running something pretty close to an EGEE/LCG/EDG supported OS.
> For a number of reasons including its external visibility, I'd class the
> VO-box as a service node, rather than a WN.
>
>
> On Wed, 12 Oct 2005, Jeff Templon wrote:
> ...
>
>>The SITE determines the OS unless special arrangements are made with the site.
>>These special arrangements are indeed *special* meaning the default OS for the
>>WNs at the site is what you should expect to be getting.
>
>
> This seems wrong: surely the VO should expect the default OS for the
> *service nodes* (CE,SE, etc.) at the site. ([Un]helpfully, this shouldn't
> be published by sites for security reasons...)
>
> If a site has its workers on a private LAN, then it might reasonably have
> say RH7.3 on the WNs but SLC on the (outward facing) service nodes (some
> sites HAVE done exactly this, IIRC).
> In that situation though, the odds of being allowed to deploy a VO-box on
> RH7.3 (or of VOs wanting to, of course!) seem pretty slim!
>
> i.e. I would expect that the default VO-box OS at a site would be the one
> already accepted there for use by externally visible machines, e.g. CE, SE
> etc.
>
> I think this still satisfies the criterion of
>
>>"let's try and get the maximum extra benefit for the VOs and at the same
>>time create as little extra work for the sites as possible."
>
>
>
> While thinking about private LANs, it strikes me that one question might
> be whether or not a VO service could run on a multi-homed machine, rather
> than all WN to VO-box traffic having to go through some gateway.
>
>
> Thanks
>
> Henry
>
|